Open source en open standaarden

In our capacity as accredited observers of the WIPO Standing Committee on Copyright and Related Rights (SCCR), we are attending the 45th session of the Committee, which is currently taking place in Geneva (April 15-19, 2024).

We made the following statement regarding limitations and exceptions for educational and research institutions and for persons with other disabilities (Agenda Item 6):

Dear Delegates,

Many of us here today will remind you that knowledge institutions face many challenges when it comes to fulfilling their public interest missions in the digital environment. These hurdles range from lack of harmonisation of copyright exceptions to legal uncertainty and fear of litigation.

In the words of Marcin, a researcher from Poland researching ancient Chinese literature and contemporary culture, and I quote “a considerable part of the work is thinking about what I can do and what I can’t do, what is legal, what is illegal”.

These obstacles are particularly problematic in a cross-border environment, where a fragmented legal framework negatively affects these activities, forcing for instance researchers to limit or abandon collaborative projects, or to select research partners according to their national copyright laws. The 2nd edition of our publication “Nobody puts research in a cage”, where we interview researchers engaged in joint and cross-border projects, shows this very clearly.

If you want case studies to understand what are the kinds of problems that you should be fixing right now, this is a good start. From researchers stuck in cages in Sweden, to researchers flying across continents to be able to research Chinese movies from the TVs of their hotel rooms, it’s unsettling to read about the obstacles they face to conduct their research projects. But it’s also fascinating to see the solutions that they propose to tackle these problems.

Sure enough, they all want more copyright exceptions, more legislation granting them rights to use copyrighted works, particularly in an international environment. And this Committee knows that there are various binding and non-binding ways of getting close to that place. With all due respect, toolkits published on an obscure corner of the WIPO website, where there are about 6000 entries for the word “toolkits”, are just not it.

We understand why this would be a priority for the Secretariat, but if this Committee is truly committed to implement their work program on L&Es, the way forward are the working groups foreseen there. And again, with all due respect, we are appalled to see that, one year after the approval of the work program, you have not been able to agree on the scope and modalities of such working groups. We thus urge you to not leave this meeting without an implementation agreement in place.

The post SCCR/45: COMMUNIA Statement on Limitations and Exceptions appeared first on COMMUNIA Association.

In our capacity as accredited observers of the WIPO Standing Committee on Copyright and Related Rights (SCCR), we are attending the 45th session of the Committee, which is currently taking place in Geneva (April 15-19, 2024).

We made the following statement regarding the protection of Broadcasting Organizations (Agenda Item 4):

Dear Delegates,

The proposed broadcast treaty, in its current version, remains a threat to the Public Domain and usage rights, particularly when legal protection of broadcasters is shaped in the form of exclusive rights, on top of rights that apply to content.

The rights-based model suggests that broadcasters will benefit from secondary rights for exploitation and control following fixation of the broadcast signal, without sufficient consideration for the public interest needs related to access to knowledge and information of signal content. In the current text, none of the exceptions are mandatory and there is no Public Domain safeguard.

Broadcasters own extensive collections of exclusive content that is highly valuable for researchers, educators, learners, cultural heritage institutions, and the general public. These collections document not only popular culture and the entertainment industry, but also function as historical documents, educational resources and research sources.

Often, the only way of accessing high-quality copies of the content in those collections is through broadcasting. Therefore, it is essential to limit exclusive rights with adequate usage rights, and ensure that, when the signal content is in the Public Domain, broadcasters are prevented from claiming exclusive rights and taking that content out of the Public Domain.

Countries opting for a rights-based model should be required to implement at least those exceptions that are already mandatory for copyrighted works (quotation, news of the day, and providing access for the visually impaired). Furthermore, they shall be required to provide in their domestic laws that, when the term of protection of the signal content has expired, the rights and protection guaranteed in this Treaty shall not apply.

The post SCCR/45: COMMUNIA Statement on Broadcasting Organizations appeared first on COMMUNIA Association.

Published research outputs often end up locked behind paywalls, unavailable to many researchers and the broader public, impeding scientific – and human – progress. Despite progress in the area of open science, research funded by the European public is no exception to this. In many cases, European taxpayers are essentially asked to pay up twice, once for funding the research and again for access to the final publication. One reason for this is the transfer of copyright ownership. To comply with contractual demands, researchers routinely outright transfer their economic exploitation rights to publishers, or fail to retain sufficient rights that would allow them or their funders to republish or reuse their work. This contradicts the primary goal of research, which is to maximise its impact by sharing it as widely as possible in a timely manner.

Today, COMMUNIA is releasing Policy Paper #17 on access to publicly funded research (also available as a PDF file), in which we propose a targeted intervention in European copyright law to improve access to publicly funded research:

We recommend a three-tiered approach to open publicly funded research outputs to the public, immediately upon publication, where a secondary publication obligation co-exists with a secondary publication right. We consider that an obligation by the funding recipients to republish is a more consequential approach to protect the public interest, as it makes Open Access (OA) mandatory, ultimately ensuring that publicly funded research outputs are republished in OA repositories. A right is, however, necessary to ensure that the authors, and subsequently the funding recipients, retain the rights necessary to comply with the obligation. A  right also provides a legal framework for the dissemination in OA repositories of publicly funded research outputs published before the entry into force of a secondary publication obligation.

In addition, we recommend the introduction of a copyright exception for the benefit of knowledge institutions, such as libraries and archives, to further support the task of making available research outputs published before the entry into force of secondary publication rights and obligations.

Resolving these issues would ideally be part of a more comprehensive regulation, a Digital Knowledge Act, which addresses the needs of research organisations and other knowledge institutions in the digital environment more broadly.

The post New policy paper on access to publicly funded research appeared first on COMMUNIA Association.

On the 6th of March 2024, we were delighted to welcome a full house at our “Unlocking Knowledge Conference” at Town Hall Europe. At this event, we asked academics, civil society representatives and EU policymakers about the issues knowledge institutions and researchers still face when accessing and using copyrighted materials in their activities, particularly in their digital activities. COMMUNIA defends that the introduction of a targeted legislative instrument, a Digital Knowledge Act for Europe, covering the needs of knowledge institutions in the digital environment, would be the way forward for aligning and updating the legal framework for European research and other public interest activities. If you were unable to make it to the event or would like to rewatch, the recording is now available.

Teresa Hackett (EIFL) moderated the first panel on legal uncertainty and exposure to liability when working with copyrighted materials. Martin Senftleben (Professor, University of Amsterdam) kicked off the panel by questioning whether the present EU legal framework is sufficiently broad for all contemporary research activities. He described the sources of legal uncertainty in the current legal framework and explained how the application of the three-step test, with its open-ended norms subject to the CJEU interpretation, gives rise to a liability risk for users. To counterbalance this risk, he highlighted that in the CJEU jurisprudence the subjective knowledge/intention of the user have been factored in the equation. He also referred to the liability limitations that exist in some civil law systems in the Americas and in the U.S. (where damages are capped when the user acted in good faith), as a potential policy mechanism that can be explored in Europe to address the liability risk.  

Annabelle Shaw (British Film Institute) provided some insight on the issues which collection managing institutions face on a day-to-day basis and made the point that smaller institutions in particular lack the legal framework to make them confident enough to make their collections accessible without fear of litigation.

Michael Arentoft (Head of Unit, European Commission DG RTD.A.4) reiterated that open science is fundamental to the EU and it is a priority action point in the European Research Area (ERA) Policy Agenda, noting the link between the European Research Area and the free circulation of knowledge in the Treaty on the Functioning of the EU (see Article 179). He highlighted the need for a system for disseminating findings and data and stated that we should take seriously the opportunity cost we are already incurring by not making datasets FAIR. He further explained some of the policy options explored in the study launched in the context of Action 2 of ERA. These options include the introduction of a general research exception and of a EU-wide secondary publication right. As a next step after the publication of the Commission’s upcoming study, the Commission will have to clarify which recommendations will lead to legislative measures and which ones to non-legislative measures. Responding to the question whether legislative measures should take the form of a regulation addressing the needs of researchers, Arentoft remarked that this would depend on the policy motivation. If they determine that there is a need for sector specific legislation, then this could take the form of a single act or a more targeted intervention, such as a secondary publication right.

Teresa Nobre (COMMUNIA) moderated the second panel on obstacles to use digital formats, due to refusals to license by rightholders, which started with a presentation of the preliminary findings of a study on this subject, commissioned by COMMUNIA and the IFLA Foundation to Christophe Geiger (Professor, Luiss Guido Carli University) and Bernd Justin Jütte (Assistant Professor, University College Dublin). Bernd Justin Jütte reminded the audience that current mechanisms to enforce users rights are insufficient and often ineffective, as right holders can still instrumentalize their exclusive rights and apply distribution (or access) strategies to prevent the proper exercise of users rights. He defended that copyright must establish positive obligations for right holders to provide access for specific purposes, arguing that an obligation to grant a specific license to enable a specific use under an exception can be derived from the ratios of the relevant exceptions.

Stephen Wyber (IFLA) presented some European examples of libraries facing refusals to license (or to license under reasonable pricing and conditions) digital materials for e-lending purposes. He encouraged us to think that the shift to digital could be seen as deregulation by stealth, urging policymakers to come up with a legislative agenda to make sure that the market as a whole also works for knowledge institutions.

MEP Karen Melchior (Renew Europe) talked about the changing role of public libraries in our current increasingly digital society. She noted that in Denmark publishers are inhibiting this transformation by not providing access to digital books at the same rate as physical ones. She warned us that, in the absence of appropriate changes to the legal framework, we might not see the expansion of access to knowledge that this digital revolution promises, and encouraged the audience to not let rights holders hijack the conversation.

In his closing remarks Felix Reda (former MEP and COMMUNIA member) made an argument for a unified act to support European research rather than surgical changes to existing legislation. Looking back on his 10 years of experience working in Parliament and civil society, he remarked that while he was sometimes able to add provisions for researchers (in case of the Digital Services Act) or damage control where necessary (Art. 17 of the EU Copyright Directive would originally have applied to University research repositories), the needs of researchers was never the starting point for any of these legislations. He reminded the audience that while universities are under pressure to operate more and more under market principles, they thus far haven’t gotten the same support to be competitive by removing bureaucratic hurdles as classical businesses have had in the past. Reda is convinced that we need a Digital Knowledge Act and asked the question: “Do we want all this public funding to be spent on actual knowledge production or on compliance with a very complicated legal mechanism?” 

The post Video recording of “Unlocking Knowledge Conference” now available appeared first on COMMUNIA Association.

The purpose of copyright, at its very basic level, finds its normative implementation in the interplay between access to protected works and the protection of the moral and material interest of creators (see Geiger, 2017). The social contract of copyright, which main purpose is to realize a broader collective concern, the access of citizens to science and culture (Geiger, 2013), lies in the approximation of the interests of rightholders and users. Unfortunately, this perspective has been long overshadowed in the EU by a traditional understanding of copyright as a system that grants exclusive ‘rights’, and under which important legitimate uses can only be performed in limited circumstances if covered by an ‘exception’. This misguided understanding of a hierarchy of rights has had a significantly negative influence on the exercise of the rights of copyright users. While rightholders enjoy broad rights to defend their interests, users must rely on exceptions with limited scope. More crucially, rightholders can instrumentalize their exclusive right to control initial access to their works, making the exercise of user rights increasingly difficult if not impossible (Kretschmer & Margoni, 2024). This is particularly true in a digital environment where access and subsequent use of protected works is further complicated by technological protection measures and paywalls.

To counter this imbalance, courts worldwide have in recent times stepped-in to clarify that ‘exceptions’ are not mere exceptions to a rule but instead define positive, enforceable rights to use copyrighted works in certain circumstances (for the EU see e.g. Funke Medien v Bundesrepublik Deutschland). A hierarchy  between rights and exceptions would be contrary to national and regional constitutional orders, in particular to the imperatives of fundamental rights (Geiger & Jütte, 2022). From this constitutional dimension of copyright emerged the notion of ‘user rights’ (Geiger, 2020).

This crucial development, which restores copyright as an access right (see Geiger, 2016; Efroni, 2010) provides a normative foundation to reinforce the societal bargain that creates incentives for authors, but also creates room for downstream creativity and innovation. However, it is only a first step towards making user’s rights a reality. Looking forward, it is essential to ensure that legitimate uses expressly foreseen by copyright law are enabled through better access to protected works. Here, it is the legal framework that is lagging behind, in particular in creating a congruence of lawful uses in analogue and digital environments. For this purpose, copyright must establish concrete obligations for rightholders to provide access for specific purposes. In the absence of positive norms, courts can support the exercise of user rights through an access-friendly interpretation of the existing framework. Spelling out these positive obligations should, however, also be a priority for future legislative reform.

Particularly in relation to digital embodiments of works and due to the changing nature of content consumption (digital access via tethered devices, streaming and cloud services), access has become easier and more convenient. At the same time, rightholders have tools and can develop strategies to tailor ‘spheres of access’. Not only can rightholders prevent certain uses by employing technological protection measures, rightholders can choose to whom, or to which institutions they grant digital access rights (i.e. licenses for specific uses). Rightholders can, for example, choose whether to license an e-book or an e-journal to a library, or whether to provide licenses to educational establishment to play music and audio-visual works for educational purposes. However, making available books or articles in digital format in institutions such as public and university libraries is needed to enable access to works, which is a necessary precondition to exercise user rights: illustrating teaching with images and sounds is itself protected by an exception (Priora, Jütte & Mezei, 2022), and researchers rely on access to information in a multitude of formats. For example, in order to text and data mine scientific publications, a researcher needs to have access to these copyrighted works, which  can be restricted by licensing and pricing policies of publishers (as only very few institutions are able to afford broad and comprehensive licenses to access to digital works across all areas of science). Added to the already established arsenal of technological protection measures and restrictive terms in online contracts, rightholders can therefore control initial access points and thereby decide whether users can benefit from their rights. The purpose of copyright, its social and innovation function, is thereby seriously undermined.

To some extent, copyright already offers mechanisms that can serve as shields against private ordering. However, these are limited to cases in which the user already enjoys lawful access to a specific work. For example, art. 6(4) of the InfoSoc Directive obliges Member States to ‘take appropriate measures to ensure’ that users of specific exceptions can exercise these if technological protection measures (TPMs) applied by rightholders prevent such exercise. This obligation is extended to the exceptions introduced by the CDSM Directive by virtue of art. 7(2). Another mechanism to ensure the proper exercise of user rights is contained in art. 7(1) CDSM Directive which prohibits the contractual override of the exceptions for non-commercial text and data mining performed by research organizations and cultural heritage institutions, for digital and cross-border teaching activities and for the preservation of cultural heritage.

In the context of the moderation practices of online platforms of the activities of their users, art. 17(9) CDSM Directive even goes one step further, explicitly recognizing for users a ‘right to exercise exceptions’ (Geiger & Jütte, 2021). To ensure that users can exercise their rights, specific online platforms are required to make users aware of this right by including a relevant reference in their terms of use. Furthermore, procedural safeguards (Quintais et al., 2019) must be established by platforms with the right to users to have recourse to the ordinary courts or other judicial authorities to ‘assert the use of an exception or limitation’.

These legal mechanisms, which are effectively tools to ensure the exercise of user rights and to shield them from the overreach of rightholders, are well-intended, but in practice difficult to enforce. Moreover, most EU member States have failed to implement express reference to these access-friendly provisions in their national laws.

But even if these user safeguards (prohibition of contractual override as an ex-ante measure and lawful access to content protected by TPMs) were properly implemented, rightholders can still apply distribution (or access) strategies to prevent the proper exercise of user rights. At least implicitly, most of the exceptions which enjoy protection under art. 6(4) of the InfoSoc Directive and 7(1) CDSM Directive require lawful access. A general ‘lawfulness’ (of access) requirement hangs the Sword of Damocles over the heads of users. In certain situations, it may not be easy to determine if the content used in exercise of a user right has been made available lawfully. For example, arts 3 and 4 CDSM Directive require lawful access as a condition for the lawful performance of text and data mining. Already the ambiguity to the ‘lawfulness’ requirement can constitute an effective barrier to the exercise of user rights (Geiger, Frosio and Bulayenko, 2019; Margoni, 2022).

Even more restrictive effects can unfold when certain uses are not subject to an exception but to a full exclusive right. For example, public lending is, by default, subject to an exclusive right under art. 2 of the Rental and Lending Rights Directive. While Member States can derogate from this right and establish and remunerated exception under art. 6(1), and even exclude certain institutions from the remuneration requirement (art. 6(3)), rightholders retain full control over the availability and modalities of licenses for public lending. Even art. 6 can require that libraries must have obtained prior lawful access in the sense of a prior transfer of ownership (VOB v Stichting Leenrecht). However, public lending through publicly funded institutions is essential for vulnerable and marginalized groups and their ability to access information and participate in cultural life, and to exercise exceptions to copyright for the purpose of self-study. While there is a good argument to be made that rightholders should be obliged to offer appropriate licenses, a stronger argument can be made that the imperatives of fundamental rights suggest that the public lending right should instead be understood, and legislatively anchored as a user right, viz. an exception or, according to art 6(1) at least a statutory remuneration right (Geiger & Bulayenko, 2022). More radically, the right to control the public lending of certain works could be eliminated (IFLA, 2016). At a minimum, in the absence of immediate legislative intervention, rightholders should be obliged to offer licenses that cater to the needs of public lending institutions on their mission to provide access to diverse collections of information.

An obligation to grant a specific license for a specific use, or rather to enable a specific use under an exception, can be derived from the ratios of the relevant exceptions – and these would include de lege lata at least those mentioned expressly in art. 6(4) InfoSoc Directive and those shielded from contractual override under art. 7(1) and art. 17(7) CDSM Directive. The purpose of these exceptions lies at the heart of copyright’s raison d’être, to create and provide access to information that will generate creativity and innovation. These objectives are also reflective of fundamental rights as protected under the EU Charter of Fundamental Rights and the European Convention on Human Rights. Chief amongst them is the right to freedom of expression, including the right to receive information, the right to education as well as the freedom of the arts and sciences, and as its derivate the right to research (Geiger & Jütte, 2021).

However, fundamental rights provide convincing arguments that access should be facilitated for certain users and privileged user groups. At a minimum, access should be facilitated for institutions which allow the realization of user’s rights: for example, libraries, public archives, educational establishments and research organizations, as well as other public interest institutions must be put in a position to function as access points for a broader public and thus to enable user’s rights. In particular, in cases where rightholders prevent digital access to specific works, which are necessary to undertake a use privileged by an exception that derive their existence from fundamental rights rationale, an obligation to provide access can be construed directly out of these fundamental rights. This obligation must, however, be subject to the condition that the work in question is already lawfully available, albeit under different conditions and possibly in different formats. In such a scenario, the relevant rightholder would be obliged to offer access to the protected work on reasonable terms.

If such an obligation were not to exist, rightholders could render user’s rights absolutely ineffective. Through contractual and technological conditions, as part of their digital distribution systems, publishers and other rightholders can tailor and segment the relevant markets around areas in which copyright exceptions normally apply. The privileging uses of protected works offered by so called ‘exceptions’ thereby fall victim to private ordering and generate societal costs of a copyright system deprived of the balancing effects of exceptions limitations (European Copyright Society, 2017). Such an obligation that directly derives from fundamental rights would be subject to the principle of proportionality, to avoid that rightholders face excessive ‘obligations to license’ and would have to incur the resulting transaction costs.

An institutional framework that administers access to works for specific purposes, and potentially supports the negotiations for reasonable and fair licensing arrangements, would be instrumental in ‘administering proportionality’ for an access right under copyright law. An EU copyright institution (Geiger & Mangal, 2022, also Frosio & Geiger, 2021) equipped with quasi-judicial competences could be key in resolving negotiation deadlocks between users and rightholders and in ensuring the exercise of user rights in practice. Such an institution could also negotiate best practices for providing access (either by suitable licenses, or by facilitating the provision of appropriate format copies) and help to formulate proposals for legislative reforms. It could also be tasked with conducting studies and research on the economic parameters that determine the conditions attached to an obligation to provide access. For example, fair conditions for access could, in terms of value and process, be determined by standards inspired from FRAND-obligations in patent law or the essential facilities doctrine in competition law. Furthermore, it could reflect and provide policy guidance when remuneration-based limitation systems ensuring access for research, teaching and cultural participation would be a workable solution to the access-problem (Geiger, Schönherr & Jütte, 2024).

Without access to copyrighted work there can be no creativity and where there is no creativity, the social contract behind copyright is broken. “With great powers comes great responsibilities”, as the biblical saying goes. It is time to enforce this social bargain so that copyright can benefit creators and the broader society at the same time and to make the EU a vibrant space for cultural participation and creativity (European Copyright Society, 2023).

This post summarizes the preliminary findings of a study conducted by the authors, which has been commissioned by COMMUNIA and the IFLA Foundation.

The post Copyright as an Access Right: Concretizing Positive Obligations for Rightholders to Ensure the Exercise of User Rights appeared first on COMMUNIA Association.

A fair and modern copyright framework is essential for an enabling European research environment. However, researchers often find that their access, use and reuse of resources essential to their activities is still limited by copyright and related rights.

While, thanks to the 2019 DSM directive, they now benefit from a EU wide text and data mining exception, the exception only covers acts of copying for a specific subset of research activities. The right of communication to the public, which is necessary for cross border or remote access to share resources and research results for the purpose of verification, validation or dissemination, is still not harmonized at the EU level. 

The 2nd edition of our publication Nobody puts research in a cage explores real life examples of the impact of the current European copyright regime on the practice of researchers working in Europe and across borders. They illustrate how the current legal framework hampers research transparency, prevents researchers from complying with open access requirements and hinders joint and cross-border initiatives.

We advocate for a EU-wide mandatory research exception to copyright and other exclusive rights for scientific research, as well as a more comprehensive regulation, a Digital Knowledge Act, addressing their needs more broadly. We also argue for international agreements on minimum standards for research to safeguard the Right to Research on a global level.

This publication by COMMUNIA, Centrum Cyfrowe, Wikimedia Sverige and the Intellectual Property Law Institute is issued as part of the Right to Research in International Copyright Law project funded by Arcadia – a charitable fund of Lisbet Rausing and Peter Baldwin.

The post 2nd edition of publication “Nobody puts research in a cage: Researchers’ perspectives on working with copyright.” appeared first on COMMUNIA Association.

Poland has been waiting for the implementation of the CDSM Directive into national law for almost 5 years. During this time some early implementation projects appeared, and in 2022 public consultations were conducted, after which things went quiet. After winning the elections in October 2023, the new ruling coalition embarked on another attempt to regulate these issues. Currently, we are at the end of yet another round of public consultations. In this blogpost we gather the most important highlights regarding the new implementation project.

In the CJEU judgement in case C-401/19, the Court ruled that, while Article 17 of the CDSM Directive is compatible with EU law, Member States have an obligation to introduce user rights protection both at the stage of content upload on platforms (ex ante protection) and after the content has been uploaded (ex post protection). The action of annulment was brought before the CJEU by Poland on the grounds that Article 17 infringed the right to freedom of expression and information, therefore we were surprised by the initial government proposal, which completely ignored the issues of the ex ante protection system, proposing only very limited mechanisms for ex post protection.

The newest implementation proposal finally seeks to address the CJEU ruling. Firstly, it points out that cooperation between online content-sharing service providers (OCSSP) and rightholders cannot lead to the unavailability of works uploaded by users in accordance with the law, especially based on provisions regarding permitted uses of copyrighted works and other subject matter. In fact, it is explicitly stated that “Actions taken by the service provider cannot lead to the automated prevention of access to works, blocking access, or their removal if the public sharing of these works does not manifestly infringe copyright.”

In addition, the online platform is required to promptly inform the user of any instance where access to the work is prevented or blocked, or if the work is removed. While this new provision is important, a more significant step in this direction could have been taken. Perhaps, following the Advocate General’s opinion in the Polish case, it would be worthwhile to explicitly state that blocking must be “limited to content which is ‘identical’ or ‘equivalent’ to works and other protected subject matter identified by rightholders.”1 Centrum Cyfrowe provided a detailed proposal for such a solution in the previous consultations. Moreover, the proposed ex post protection system by the drafters also requires refinement and significantly diverges from the German approach in this matter.

Unfortunately, the new proposal also includes a number of provisions that are problematic. Let’s examine a few of them.

First, the Polish government proposes that the implementation of provisions regarding Text and Data Mining (TDM) into Polish law shall exclude the training of generative artificial intelligence (GAI) models. This exclusion is intended to apply to both research-oriented TDM and commercial TDM. According to the government, this approach stems from the fact that during the creation of the CDSM Directive, this type of TDM application was not considered, and GAI models were not as prevalent at the time, thus they must be excluded. As correctly pointed out by Paul Keller, this rationale is entirely flawed. Currently, it is widely recognised that the TDM provisions outlined in the CDSM Directive cover the mining of data for the purpose of training artificial intelligence. This interpretation is also supported by the European Artificial Intelligence Act (AI Act). While generative AI introduces complexities in power relations and the distribution of benefits in society and may require a systemic approach, the implementation of TDM into Polish law is certainly not the appropriate avenue for this discussion for several reasons.

To begin with, commercial TDM regulations are based on an opt-out clause. Therefore, if someone does not want their works to be used for the purpose of AI training, they can simply opt out, which is already happening. Excluding this application of TDM, even from the scientific research exception, seems entirely unjustified. The right to research is of constitutional importance and researchers worldwide are conducting studies in the field of artificial intelligence. Depriving Polish researchers of such opportunities not only appears to be unconstitutional but is also directly detrimental to the development of Polish science. If this exclusion was to be adopted, Polish researchers would not only be unable to conduct such research themselves, but the entire international cooperation of researchers from Poland with other countries would be jeopardised. This issue has already caused significant concern in the academic world, and raised demands for the government to quickly withdraw this proposal during the revision process.

Second, the proposal is not compliant with Article 8 of the CDSM Directive. The CDSM Directive introduced a system for the use of out-of-commerce works (OOCW). Importantly, this change was made in response to the practical inapplicability of the directive on orphan works. The adopted system for OOCW was intended to facilitate real access to those works that would otherwise not be accessible because no one is currently interested in bringing them to market. Under this system, cultural heritage institutions can provide access to such works either based on a licence with representative collective management organisations (CMOs) or, in the absence of such CMOs, based on a relevant exception. This dichotomous division is quite clear and allows for easy determination of whether specific works can be made available. Simply put, one only needs to check if there is a representative CMO. However, according to the Polish legislator, cultural heritage institutions will no longer only have to determine whether there is a representative CMO but also whether a given creator is not represented by any CMO. If a creator is represented by any CMO, it would be impossible to use the OOCW exception. In practice, this introduces the necessity of conducting rights clearance for each creator separately and determining whether they are represented by any CMO and to what extent. Ultimately returning to an unworkable exception, as was the case under the orphan works directive.

Third, and finally,  the issue of protection against exceptions’ contractual override and their blocking using Technological Protection Measures (TPM) remains untouched as, for no particular reason,  the Polish legislator decided not to implement Article 7 of the CSDM Directive directly. Concerning TPMs, according to the drafters, appropriate regulations were already introduced during the implementation of the InfoSoc Directive, and since they apply to the new exceptions, there would be no need to change anything in this regard. Unfortunately,  the implementation under the InfoSoc Directive is rudimentary at best and is essentially limited to the issue of liability for circumventing TPM.

When it comes to protecting exceptions against contractual override, according to the legislator, under Polish law, all exceptions have a mandatory character (ius cogens). Indeed, this also applies to those mentioned in Article 7 of the DSM Directive. While this is assumed by the “dominant” part of legal doctrine as well as the Polish legislator (as evidenced by the legislative proposal), we are not aware of any Polish court ruling that would confirm or deny such an approach. We generally agree that this should be the case, however,  in order to create sufficient legal certainty, this issue should be explicitly settled in the law. Importantly, it’s not just about meeting the requirements set out in the Directives. People and public institutions make use of provisions for fair use. The risk of copyright infringement, which can lead to liability without fault, is significant due to the complexity of the matter and the three-step test, which does not facilitate its application. Assuming that such “users” of the law will base their knowledge of its content and principles on something other than the text of the law is very problematic. After all, in the normal course of action, no one will consider what legal scholars think about this matter or what the legislator wrote in the rationale. For a long time, it has been recognised that creating clear norms regarding fair use is crucial for its practical application. As such, introducing this additional layer of legal uncertainty may well have a chilling effect on the use of works.

At this point, it is worth emphasising what is missing in this proposal: Namely, a vision for balancing the rights of rightholders and users fitting for a  regulation in the 21st century. On the one hand, it is understandable that the new government, which must urgently implement a Directive whose implementation obligation passed years ago, does not have much space to think about the entire ecosystem. On the other hand, the ministry that prepared this project should strive to introduce the best possible regulations. For a long time,  the need to search for legal instruments that would strengthen user rights was strongly identified by social organisations, in academic literature and CJEU jurisprudence. In particular, exceptions and limitations to copyright have long been recognised in CJEU case law as rights.2 However, to move from this rhetoric of rights to actual rights, concrete regulations at the national law level are needed, in particular where rightholders prevent uses of works that are in accordance with the law. Of course, in this context, it is also worth considering provisions to counter abuses of copyright.

Finally, it would be desirable to introduce regulations enabling e-lending of books by libraries based on books digitised by them. It seems that this is possible under the currently applicable EU law. The case for such a possibility is supported by the results of research conducted by the Centrum Cyfrowe in cooperation with the Jagiellonian University under an international study of digital lending regulations carried out for Knowledge Rights 21. Similarly, it would be worthwhile to use this opportunity to expand the provisions regarding the use of works for the purpose of conducting scientific research, rather than attempting to introduce a quantitative limitation on how much can be used within such activities (20%) – as currently proposed by the legislator. In this context, it would also be worthwhile to introduce the Secondary Publication Right (SPR) into Polish law. SPR are “the rights of the researchers and their employers to reuse a work after the first publication of a formal version; in this case to republish it on an  infrastructure  that  by  its  nature  (public  and  not-for-profit)  is  coherent  with  the (public) nature of the funding source.”3 Of course, it would also be worth considering the possibility of creating safe harbours for cultural heritage institutions regarding their good faith use of works for the fulfilment of their public mission, or the possibility of internet archiving by these institutions. In general, it is essential to consider an internal balance within copyright law so that the constitutional rights of rights holders are properly balanced with the constitutional rights of users. The new Policy Recommendations of COMMUNIA can serve as a constant inspiration for this.

1. OPINION OF ADVOCATE GENERAL,SAUGMANDSGAARD ØE delivered on 15 July 2021 (1), Case C‑401/19, paragraph 201.
2. see: JUDGMENT OF THE COURT (Grand Chamber), 29 July 2019, In Case C‑469/17, point 161.
3. Tsakonas, G., Zoutsou, K., & Perivolari, M. (2023). Secondary Publishing Rights in Europe: status, challenges & opportunities. Zenodo. https://doi.org/10.5281/zenodo.8428315, p. 18.

The post The Good, the Bad and the Missing – the new proposal for the implementation of the CDSM Directive into Polish law. appeared first on COMMUNIA Association.

Access and use of public sector documents and political speeches are key to a thriving democracy and civil society as they ensure transparency and accountability. Furthermore, they provide society with important historical records about political decision making. Copyright and paywalls, however, can render access to these materials difficult or prohibitively expensive. Today, COMMUNIA is releasing a new policy paper #16 “on public sector documents and public speeches” (also available as a PDF file), in which we are proposing EU level solutions to guarantee access to an usability of these essential documents, as well as to political and other public speeches.

The paper explores the legal basis for excluding public sector documents and public speeches from copyright both from a historical perspective citing the original intent of the Berne convention as well as by looking at contemporary legislation. While it is true that at present some EU member states individually already exclude public sector documents from copyright protection, this freedom is not clearly granted by all Member States. Public speeches, on the other hand, are subject to copyright exceptions in the vast majority of Member States, indicating a significant convergence of policy positions.

COMMUNIA therefore argues for an EU level solution, moving beyond the shortcomings of existing EU legislation, most notably the Open Data Directive. We advocate for excluding public sector documents from the scope of copyright protection altogether. For public speeches we similarly recommend exclusion from copyright or alternatively to turn the optional exceptions in Article 5(3)(e) and (f) of the InfoSoc Directive into mandatory exceptions. Resolving these issues would ideally be part of a more comprehensive regulation, a Digital Knowledge Act, which addresses the needs of universities and other knowledge institutions, such as schools and cultural heritage institutions, in the digital environment more broadly.

Note that we left aside publicly funded research, which we will address in our next policy paper.

The post Access to and use of public sector documents and public speeches – new policy paper appeared first on COMMUNIA Association.

This article was first published on the Kluwer Copyright Blog on date 20/02/2024

When life gives you lemons, make lemonade. This must have been the key insight at the Polish Culture and National Heritage Ministry when the new administration took over and discovered that more than 2.5 years after the implementation deadline, Poland still had to implement the provisions of the 2019 Copyright in the Digital Single Market Directive into national law. So how do you make lemonade out of the fact that you are the only EU Member State without an implementation? You claim that the delay allows you to propose a better implementation.

In this particular case, the government claims that the delay allowed it to properly consider the impact of generative AI on copyright and come to the conclusion that training generative AI systems on copyrighted works does not in fact fall within the scope of the text and data mining exceptions contained in the directive. From the explanatory memorandum accompanying the draft implementation law published on Thursday last week for public consultation (all quotes below are own translations from the Polish original):

The implementation of the directive now, in 2024, dictates that we refer here to the issue of artificial intelligence and the question of whether text and data mining within the meaning of the directive also includes the possibility of reproducing works for the purpose of machine learning. Undoubtedly, at the time the directive was adopted in 2019, the capabilities of artificial intelligence were not as recognizable as they are today, when “works” with artistic and commercial value comparable to real works, i.e., man-made, are beginning to be created with the help of this technology. Thus, it seems fair to assume that this type of permitted use was not conceived for artificial intelligence. An explicit clarification is therefore introduced that the reproduction of works for text and data mining cannot be used to create generative models of artificial intelligence.

This “explicit clarification” can be found in the text of the proposed implementation for both articles 3 and 4 of the CDSM directive. The article 3 implementation states that cultural heritage institutions and academic research organizations…

may reproduce works for the purpose of text and data mining for scientific research, with the exception of the creation of generative models of artificial intelligence, if these activities are not performed for direct or indirect financial gain.

The same exception to the exception can also be found in the implementation of the general text and data mining exception:

It is allowed to reproduce distributed works for the purpose of text and data mining, except for the creation of generative artificial intelligence models, unless otherwise stipulated by the authorized party.

It is worth stressing that the language quoted above is contained in the public consultation version of the implementation law and thus not final. It also seems clear that this language has not been widely consulted within the Polish government as it clearly contradicts efforts undertaken by other parts of the government. Still it is worth taking a closer look at the rationale behind this implementation and to assess the conformity with the provisions of the directive and the overall impact of the proposed approach.

First of all, while it is understandable that lawmakers seek more clarity about the relationship between the EU copyright framework and the use of copyrighted works for training AI models, the assumption that the TDM exceptions were “not conceived for artificial intelligence” is simply wrong. While there is little publicly available documentation of what lawmakers had in mind when they agreed on the structure of the TDM exceptions, what is available makes it clear that the development of artificial intelligence was explicitly factored into the discussions. Both the European Parliament statement and the European Commission’s explainer of the directive, published after the adoption of the directive in March 2019 specifically highlight that the TDM exception in Article 4 was introduced “in order to contribute to the development of data analytics and artificial intelligence”.

If there was any doubt if the exception was conceived in order to facilitate the development of generative Artificial Intelligence, this relationship was further clarified in March 2023 (at a time when the impact of Generative AI was widely recognized). In response to a Parliamentary question that suggested that “The [CDSM] Directive does not address this particular matter”, Commissioner Breton pointed out that TDM exceptions do in fact “provide balance between the protection of rightholders including artists and the facilitation of TDM, including by AI developers”.

Finally the upcoming Artificial Intelligence Act — which has been supported by the Polish government — contains a provision that points out that developers of generative AI systems must “put in place a policy to respect Union copyright law in particular to identify and respect, including through state of the art technologies, the reservations of rights expressed pursuant to Article 4(3) of [the CDSM] Directive”. In addition, the AI act also contains a recital (60i) that explains the interaction between the training of generative Ai systems and the exceptions contained in article 3 & 4 of the copyright directive.

All of this makes it clear that “now, in 2024” the TDM exceptions as introduced in 2019 do in fact provide the framework for the use of copyrighted works for the purpose of training generative AI systems, even though some stakeholders would much prefer that this was not the case.

It is also clear that any attempt to exclude from the scope of the TDM provision the reproductions made in the context of training generative AI models would, prima facie, result in a non-compliant implementation. Defined in Article 2(2) as “any automated analytical technique aimed at analyzing text and data in digital form in order to generate information which includes, but is not limited to, patterns, trends and correlations”, the term must be considered as an autonomous concept of EU law that cannot be modified by Member States in line with political considerations. As outlined above, there is a broad consensus that the concept of text and data mining includes the training of AI models. Even if the Polish Ministry of Culture and National Heritage does not wish this to be the case, it must still implement the Directive without changing a core concept introduced in the Directive.

While we are waiting for the consultation process to play out, it is instructive to consider what would be the consequences should the TDM exception be implemented as proposed by the Ministry of Culture and National Heritage. By excluding the creation of generative artificial intelligence from the scope of both TDM exceptions, the Polish copyright law would remove any statutory basis for the use of copyrighted works in the context of building generative AI models. This would require AI developers to obtain permission from all rightsholders whose works are included in their training data. Given the amounts of copyrighted works that are required to train the current generation of AI models (often measuring in the billions of individual works) this would likely be impossible for anyone but the most well-resourced companies making it virtually impossible for smaller companies or public efforts (such as the Polish open PLLuM language model), as they would lack the resources to undertake the effort required to obtain the required permissions.

What is especially stunning in the Polish implementation proposal is that it not only excludes the creation of AI models from the scope of the Article 4 exception (which applies to commercial AI developers) but also from the scope of the Article 3 exception (which is designed to enable non-profit scientific research) which seems especially short sighted. The implementation proposal should therefore be read like a misguided attempt to hinder any development or use of generative AI models in Poland.

At this point, it seems useful to recall the key balances inherent in the EU’s regulatory framework for the use of copyrighted works in AI training. They form the basis of claims by the Commission and others that the EU has a uniquely balanced approach to this thorny issue. Taken together, the TDM provisions address 4 key concerns: (1) They limit permission to use copyrighted works for training data to those works that are lawfully accessible. They (2) privilege non-profit scientific research, (3) they ensure that creators and other rights holders can exclude their works from being used to train generative AI systems, and (4) they ensure that works that are not actively managed by their rights holders can be used to train AI models.

Excluding the training of generative AI from this balanced arrangement may please some creators and rights holders, but it also pushes AI back into a legal gray area. It also seems incompatible with the provisions of the AI Act, which situates the training of generative AI models within the broader concept of TDM, and which will be directly applicable in Poland.

What is needed, instead of efforts to undermine the existing framework, are measures to ensure that the current approach can work in practice. The new copyright provisions in the AI act are an important step into this direction, but they need to be complemented by the creation of a public infrastructure to facilitate opt-outs and measures aimed at ensuring fair licensing arrangements between rights holders and AI developers.

The post TDM: Poland challenges the rule of EU copyright law appeared first on COMMUNIA Association.

This is the fourth blog post in our series on trends in national copyright policy following the DSM implementation. With the national processes finalised in all but one Member State, we’re evaluating the transpositions to understand the impact of our proposals, identify areas of convergence and divergence, and assess how those affect the position of users and the Public Domain. Previously, we presented our research on the implementation of research rights, education rights, the Public Domain safeguard and cultural heritage rights. In this blog post, we share the results of our comparative analysis on the implementation of the new press publishers’ right.

Article 15 of the DSM Directive establishes a new related right for press publishers, allowing them to control online uses of their press publications by information society service providers. More specifically, Member States are required to grant to press publishers established in the EU the exclusive rights to authorise or prohibit the reproduction and making available to the public of their publications by information society service providers. As a result, any online uses of press publications by those service providers that go beyond individual words and very short extracts can only take place with the press publishers’ permission.

While Article 15 targets uses by information society providers, it negatively affects access to information by the wider public within the EU. Therefore, our priority during the national transpositions was to ensure that all legal limits to this new right were respected. For this reason, our comparative analysis focuses only on the implementation of the definition of press publications in Article 2(4), on the exclusions from the scope of protection in Article 15(1), on the protection of uses of public domain publications and uses permitted under non-exclusive licences in Article 15(2) in fine, and on the optional and mandatory limitations and exceptions as permitted by Article 15(3). For an examination of other elements of Article 15, we recommend this study by Christina Angelopoulos.

Nearly all elements of Article 15 are prescriptive. Yet that did not stop several Member States from leaving their own mark during the implementation of this provision. Starting with the definition of press publication, several national lawmakers introduced changes to the text even where the EU law did not afford them that flexibility.

Most of the variations encountered in the definition of press publication are probably innocuous. For instance, three Member States (Cyprus, Malta and Spain) introduced language from recital 56, which clarifies that the new right shall not apply to “websites, such as blogs, that provide information as part of an activity that is not carried out under the initiative, editorial responsibility and control of a service provider, such as a news publisher.” We do not see any conflict with the text of the Directive in those instances and, in our recommendations on Article 15, we even suggested the inclusion of that example in national laws, to make the subject matter clearer.

In a few cases, however, the changes to the definition might have an impact on the delimitation of the press publishers’ rights. For example, Article 2(4) in fine carves-out “periodicals that are published for scientific or academic purposes” from the definition of press publication, but Austria extended the carve-out to periodicals published for artistic purposes, offering literary magazines as an example of such a periodical. More concerning from our perspective, however, are modifications or omissions that might result in expanding the subject matter covered by the new right. To give a couple of examples: Denmark does not have a definition of press publication in the law, and Austria failed to implement Article 2(4)(c), which requires that the press publication must be published in any media under the initiative, editorial responsibility and control of a service provider.

Still, non-compliant implementations of Article 2(4) are far from being the norm: aside from Denmark, only six Member States have a definition of press publication that does not seem to fully comply with the Directive. The remaining nineteen Member States either follow closely the wording of the Directive or introduce changes to the definition that are apparently inconsequential:

• Definition of press publication (probably) complies with EU law: BE, BG, CY, CZ, FI, DE, GR, HU, IE, IT, LV, LU, MT, NL, PT, RO, SK, SL, SE
• Definition of press publication (probably) not fully complies with EU law: AT, HR, EE, FR, LT, ES
• No definition of press publication: DK

From a users’ rights perspective, the exclusions in Article 15(1) are extremely important. They offer a negative delimitation of the scope of protection of the new right, preventing its expansion to uses that are not made by information society providers (private or non-commercial uses of press publications by individual users), acts that are not covered by copyright (acts of hyperlinking), uses that cannot be appropriated by anyone (uses of individual words) and uses that are protected by fundamental freedoms (uses of very short extracts of a press publication). It is thus very concerning that only a few Member States have properly implemented all of these substantive limits.

Looking into the implementation of the second paragraph of Article 15(1), which excludes private or non-commercial uses of press publications by individual users, we verify that the vast majority (twenty-two Member States) implemented both limits. Yet, in one of those Member States (Portugal) the implementation does not fully comply with the DSM Directive, since private uses are subject to two additional conditions: the individual user must have lawful access to the press publication and the use must take place in the context of the exercise of the right to be informed. Belgium, the Czech Republic and France bypassed the implementation of this paragraph altogether, and Sweden only excluded non-commercial uses:

• Exclude private uses: AT, BG, HR, CY, DK, EE, FI, DE, GR, HU, IE, IT, LV, LT, LU, MT, NL, PT (subject to further conditions), RO, SK, SL, ES
• Do not exclude private uses: BE, CZ, FR, SE
• Exclude non-commercial uses: AT, BG, HR, CY, DK, EE, FI, DE, GR, HU, IE, IT, LV, LT, LU, MT, NL, PT, RO, SK, SL, ES, SE
• Do not exclude non-commercial uses: BE, CZ, FR

The implementation of the third paragraph of Article 15(1) is more successful. All Member States establish that the press publisher’s right shall not apply to acts of hyperlinking, and only Cyprus has narrowed down the reach of the exclusion, by applying it to links that can be “activated” by the user (thus excluding other forms of links such as those that use framing technology).

The fourth and final paragraph of Article 15(1) contains one of the main substantial limitations to the new press publishers right, the exclusion of uses of individual words or very short extracts of a press publication. All Member States implemented this provision, but not all stuck to the wording of the Directive. In spite of the fact that the EU lawmaker intended the notion of “very short extracts” to be an autonomous concept of EU law, subject to a harmonised interpretation, three national laws limit the length of such extracts. Six Member States, on the other hand, added a new condition borrowed from recital 58, which states that “the exclusion of very short extracts (must) be interpreted in such a way as not to affect the effectiveness of the rights provided for in this Directive”:

• Use of very short extracts not subject to further conditions: AT, BE, CY, CZ, DK, EE, FI, DE, HU, IE, LV, LU, MT, NL, PT, SL, SE
• Use of very short extracts subject to quantitative restrictions: BG, LT, RO
• Use of very short extracts subject to the condition that the use must not affect the effectiveness of the right and/or be capable of replacing the press publication and/or be capable of exempting the reader from referring to it: HR, FR, GR, IT, SK, ES

Two other important restrictions to this new right are set forth in the last two sentences of Article 15(2). According to these provisions, the new exclusive right may neither be invoked to prohibit uses by other authorised users (on the basis of a non-exclusive licence) nor to prohibit uses of works or other subject matter for which protection has expired.

As an organisation that works to defend the Public Domain from misappropriation and that strongly supports the use of open licences to improve access to knowledge, we would have expected to see these mandatory restrictions implemented in all Member States. However, some Member States have missed out the implementation of one or both of these provisions:

• Protects uses by other authorised users: AT, BG, CY, FI, FR, DE, GR, HU, IE, IT, LV, LT, LU, MT, NL, PT, RO, SK, SL, ES
• Does not protect uses by other authorised users: BE, HR, CZ, DK, EE, FR, HU, SE
• Protects uses of Public Domain publications: AT, BG, CY, FI, DE, GR, HU, IE, IT, LV, LT, LU, MT, NL, PT, RO, SK, SL, ES
• Does not protect uses of Public Domain publications: HR, CZ, DK, EE, FR, HU, SE

Finally, Article 15(3) states that the new right available to press publishers is subject to all the optional and mandatory limitations and exceptions laid out in the InfoSoc Directive (including those allowing quotations), in the Orphan Works Directive, in the Directive implementing the Marrakesh Treaty and in the DSM Directive itself. A detailed comparison of the exceptions applicable to the press publisher right and other related rights reveals, however, that in a few cases Member States decided not to extend all of those relevant exceptions to the new right:

• Press publishers’ right subject to all national L&Es applicable to other related rights: AT, BE, BG, CY, DK, EE, FR, DE, GR, HU, IT, LV, LT, LU, MT, NL, PT, RO, SK, SL, ES
• Press publishers’ right not subject to all national L&Es applicable to other related rights: HR, CZ, FI, IE, SE

Article 15 is problematic, from COMMUNIA’s perspective, because it creates a new layer of exclusive rights on top of copyright (journalists and press publishers continue to benefit from copyright protection with regards to their original works and publications), without a clear demonstrated need or economic evidence that such exclusivity was needed.

The objective of the EU lawmaker was to improve the downward-sliding financial outlook of press publishers in the online environment, but an exclusive right does not necessarily translate into revenues. So much so that a few national lawmakers decided to go a step further, introducing licensing provisions, or even resorting to competition law to force information society service providers to licence content from press publishers for reuse on their news aggregators and media services.

In spite of that, several Member States ended up with a new exclusive right that has a broader scope of protection than what was intended by the EU lawmaker. Out of the twenty-six Member States analysed, only eight make a compliant implementation of all the mandatory limits foreseen in the Directive. A poor result that should definitely deserve the Commission’s attention.

In the final blog post in this series, we’ll focus on the implementation of Article 17.

The post The Post-DSM Copyright Report: the press publishers’ right appeared first on COMMUNIA Association.

This is the third entry to a series of blog posts highlighting the positive and negative trends in national copyright policy following the DSM implementation. With the implementation process finalised in 26 out of 27 Member States, we’re comparing the national implementations to understand the impact of our proposals, identify areas of convergence and divergence, and evaluate how those affect the position of users and the Public Domain. Previously, we presented our analysis on research rights and education rights. Today, we share the results of our study on the implementation of the Public Domain (PD) safeguard in Article 14, together with some highlights from our partial scanning of the legal landscape for cultural heritage rights.

Article 14 of the DSM Directive regulates the legal status of reproductions of public domain works of visual art. This provision requires Member States to exclude faithful reproductions of those works from the scope of protection of copyright and related rights, ensuring that they cannot be taken out of the Public Domain. This is the first EU legislative intervention aimed at protecting the Public Domain from misappropriation, and targets the practice of museums in claiming exclusive rights over materials resulting from digitizations of public domain works in their collections.

Articles 6 and 8, by contrast, grant certain rights to cultural heritage institutions (such as museums, libraries and archives) over their collections. Article 6 introduces a mandatory exception to copyright, authorising cultural heritage institutions to take copies of works in their collections for preservation purposes. Article 8 puts in place a licensing mechanism and a fallback exception (applicable when there is no representative collective management organisation to issue licences for certain uses and categories of works), intended to allow cultural heritage institutions to digitise and make out-of-commerce works in their collections available online.

Article 14 is directed towards Member States that afford exclusive rights protection to non-original photographs. While a photograph that does not fulfil the originality criteria cannot be protected under EU copyright law, a small number of Member States use the policy space available to them under Article 6 of the Term Directive (which allows Member States to protect “other photographs” at the national level) and grant related rights over photographs that do not qualify as photographic works.

Whereas only seven Member States (out of the 26 analysed) have a double system of protection of photographs, a total of seventeen Member States ended up transposing Article 14:

• Protects non-original photographs, implements Article 14: AT, DK, FI, DE, IT, ES, SE
• Protects only original photographs, implements Article 14: BG, HR, CY, EE, GR, LV, LT, MT, PT, RO
• Protects only original photographs, does not implement Article 14: BE, CZ, FR, HU, IE, LU, NL, SK, SL

In Member States that do not grant protection to non-original photographs, the scope of application of Article 14 appears to be very limited, as nearly all the materials resulting from faithful reproductions of works of visual arts are already free from related rights. Indeed, digitizations of works of visual arts in the Public Domain are typically made through non-original photography or by other means that do not give rise to related rights protection (e.g. scanning). Sure enough, where film is the means chosen to capture the work of visual art, the materials resulting from the reproduction can also give rise to related rights (the rights of the film producer), but in those cases chances are that the resulting material also consists of an original work protected by copyright, in which case Article 14 no longer applies.

In that sense, it is understandable that so many national governments decided not to implement Article 14. The Public Domain safeguard only deals with a specific category of works (works of visual art). Abstaining from transposing this provision, as it is, avoids bringing into those legal systems unnecessary ambiguity, stemming from the fact that faithful reproductions of all categories of works are excluded from copyright and related rights, but only those pertaining to a specific category of works are expressly protected against appropriation and only insofar as those works have already fallen into the Public Domain.

Having said that, in some Member States (e.g. Bulgaria, Greece, Italy, Portugal, Slovenia and Spain), cultural heritage laws impose a similar type of exclusive rights over cultural heritage images held in their national collections. Therefore, in some Member States, asserting the right to use faithful reproductions of public domain works of visual arts might turn out to be more important due to the exclusivity claimed by cultural heritage laws than to the rights granted by copyright laws.

Although the DSM Directive does not expressly solve this conflict of legal regimes, it is clear that the European lawmaker wanted to free all materials resulting from faithful reproductions of works of visual arts in the Public Domain from any form of exclusivity, in order to enable the full enjoyment of those works by the general public. In this context, the Italian implementation of Article 14 is particularly controversial: It dictates that the protection imposed by cultural heritage laws prevails over the Public Domain safeguard. To add insult to injury, the Italian courts have recently issued a number of rulings asserting personality rights to cultural heritage based on Italian civil law.

The inadequate implementation of Article 14 in Italy raises many concerns, but it is far from being the only case where a Member State has decided to narrow down the Public Domain safeguard. As pointed out by Europeana, four Member States (Austria, Denmark, Finland and Spain) apply the safeguard only to works of fine arts. Furthermore, there are also divergent approaches as to the application in time of this provision, some more restrictive than others.

With Europeana and cultural heritage institutions leading the copyright advocacy efforts concerning their activities, it has not been COMMUNIA’s top priority to track down the implementation of Article 6 and 8. Nonetheless, we could not avoid noticing some problematic interpretations of the latter.

In Portugal, some of the policy choices made by the national government when implementing Article 8 clearly do not comply with the requirements of the Directive. First of all, the exception for the use of out-of-commerce works and other subject matter provided for in Article 8(2) only applies to materials published, communicated to the public or made available to the public prior to 1 January 1980. This means that, at any given point in time, the pool of out-of-commerce materials that can be used under the exception is forever limited to those materials that have been published or made available before 1980. Secondly, this cut off date only applies to the fallback exception in Article 8(2) – it does not serve to determine which materials can be licensed by collective management organisations under the licensing mechanism in Article 8(1).

In Italy, the scope of the exception in Article 8(2) has also been limited in a way that does not  seem to comply with the Directive. Here, the exception only applies to two categories of materials, software and databases. Apparently, the Italian government has come to the conclusion that, at the time of the implementation, there were sufficiently representative collective management organisations for all other categories of out-of-commerce materials. Even if that was the case, which is very unlikely, the lawmaker should have accounted for future changes in the CMOs landscape in Italy and fully implemented the fallback exception.

Overall, the transpositions of Article 8 and Article 14 across the EU seem to satisfy the requirements of the DSM Directive. However, the errors identified in a few transpositions are relevant enough for the Commission to initiate infringement proceedings against those countries.

As an organisation dedicated to defend the Public Domain, COMMUNIA is particularly apprehensive that the implementation of the Public Domain safeguard only had the intended effect in some of the targeted Member States. The fact that a few national governments decided to narrow down the scope of provision causes significant concerns. The scope of Article 14 is already quite circumscribed: it applies only to works of visual arts in the Public Domain. Restricting this category even further, to cover only works of fine arts, is revealing of the resistance that some Member States have to embrace the Public Domain. Even more telling is the Italian exclusion of national cultural heritage altogether from the scope of protection of this safeguard.

Italy has shown us clearly that the Public Domain can be appropriated not only by copyright owners, but also by national governments and institutions. That the EU copyright policies can be so easily overridden by cultural heritage and civil laws should raise some eyebrows in Brussels. While only the courts in Luxembourg can provide definitive guidance on the application of Article 14, there is certainly room for the EU lawmaker to step forward and solve the conflicts stemming from the overlaps between EU copyright laws and local civil and cultural heritage laws. Again, a Digital Knowledge Act for Europe could go a long way in lifting all legal restrictions to the enjoyment of our shared cultural heritage.

The post The Post-DSM Copyright Report: Public Domain and cultural heritage rights appeared first on COMMUNIA Association.

This is the second entry to a series of blog posts highlighting the positive and negative trends in national copyright policy following the DSM implementation. With the implementation process finalised in all but one Member State, we conducted a comparative analysis of the national implementations to understand the impact of our proposals, identify the areas where Member States have varied or converged, and evaluate whether these variations affect the position of users and the Public Domain. The first blog post presented our findings concerning the implementation of research rights. In this follow-up, we share the results of our analysis of the legal landscape for education.

Article 5 of the DSM Directive made it mandatory for Member States to introduce (or maintain)  an exception or limitation to copyright giving educators and learners at educational institutions the freedom to use copyrighted materials in digital and cross-border teaching and learning activities.

Member States had some freedom in how to implement this provision. While they had to respect the minimum scope of protection set by the DSM Directive, covering at least the same beneficiaries (educational establishment and its pupils, students and teachers), uses (“digital uses”), purposes (“illustration for teaching”) and subject matter (“works and other subject matter”), they were free to go beyond this mandatory provision. According to Article 25, Member States can adopt a broader educational exception, as long as it is compatible with existing EU law (namely the InfoSoc Directive).

However, national governments also had the option to set more conditions than those mandated by the Directive. Specifically, they could impose quantity limitations and subject the exception to compensation and licence availability.

Prior to the introduction of this mandatory exception, some Member States had strong educational exceptions while others had only narrow provisions, which oftentimes did not apply outside the physical classroom. This created inequalities in using digitally-supported education within the EU territory, and interfered with cross-border activities. With the DSM forcing a minimum set of rights across the region, those obstacles have been largely removed.

On the national level, the legal landscape for digital and distance education has significantly improved in countries with limited exceptions. The situation is less clear in countries with robust education exceptions, as most Member States implemented the DSM exception alongside their flexible InfoSoc-inspired exceptions. This results in unnecessary overlaps, which in turn makes it more difficult to assess whether those national frameworks remain as broad as before.

Be that as it may, overall the national copyright reforms prompted by the DSM Directive seem to have resulted in digital education exceptions with a broader scope of protection than before. Out of the 26 Member States analysed, only three Member States present a more limited framework for education after the implementation than they used to:

• Scope of digital education exceptions (probably) broader than before: BG, DK, FI, DE, GR, HU, IT, LV, LT, PT, SL, ES, SE
• Scope of digital education exceptions (probably) the same as before: AT, BE, CR, CZ, EE, IE, LU, NL, RO, SK
• Scope of digital education exception narrower than before: CY, FR, MT

In addition, the majority of Member States now has a legal framework for education broader in scope than the minimum standard imposed by the DSM.

Twelve Member States have educational exceptions that either extend to more acts of use (e.g. distribution), cover more users (e.g. public institutions, the educational establishment’s technical staff or any user), or apply to more purposes (e.g. research, professional training or lifelong learning) than the DSM. Six Member States have a specific digital and cross-border education exception modelled on the DSM exception and, in parallel, an open education exception that applies to all educational uses by all sorts of users, the only condition being that the use is made for non-commercial educational purposes. And one Member State introduced an open exception instead of following the DSM prototype:

• Introduces open education exception instead of DSM exception: LV
• Introduce or maintain open education exception in parallel to DSM exception: HR, CZ, EE, LU, RO, SK
• Introduce or maintain education exceptions covering more uses, users or purposes than the DSM exception: BE, HR, CY, ES, FR, DE, HU, IE, LT, MT, NL, PT

As to the conditions applicable to the new exception, the main focus of Communia during the implementation processes was to advocate against rights holders’ efforts to subject the exception to licence availability and – in Member States with a tradition of uncompensated education exceptions – oppose attempts to subject digital uses to compensation.

We recall that the DSM Directive does not require compensation for uses made under the new exception, but expressly states that Member States are free to subject such uses to the payment of a fair compensation to copyright owners, if they wish to.

Prior to the DSM, eighteen Member States had education exceptions that were completely or largely unremunerated. The DSM implementation led to a change of policy in four of those Member States: Cyprus, Greece, Malta and Slovenia introduced, for the first time, a compensation requirement to their education exceptions. By contrast, a few Member States with a tradition of compensated exceptions opted for not subjecting these new uses to compensation.

In total, seventeen Member States permit the uses covered by the DSM to take place without compensation, two require compensation only for some of those uses, and seven require it for all uses:

• Do not require compensation for uses covered by the DSM exception: BE, BG, HR, CZ, EE, FI, HU, IE, IT, LV, LT, LU, PT, RO, SK, ES, SE
• Require compensation for some uses covered by the DSM exception: DE, NL
• Require compensation for all uses covered by the DSM exception: AT, CY, DK, FR, GR, MT, SL

We fiercely opposed the possibility to give preference to licensing offers over the education exception both during the negotiations that lead to the adoption of the Directive and throughout the national implementation processes. A legal framework where a teacher or student’s ability to benefit from the exception can be taken away by copyright owners, once they start marketing licences for those uses, regardless of the amount and purpose of the use, sets a dangerous precedent for users’ rights.

Thirteen of the Member States that have so far implemented the Directive opted to not introduce this condition into their laws. The other half, however, did exactly the opposite: seven Member States subject all uses permitted under the exception to licence availability, and the other six only do so for a limited category of works (teaching materials and/or music scores):

• No uses subject to licence availability: BG, HR, CZ, EE, HU, LV, LT, LU, NL, PT, RO, SK, ES
• Some uses subject to licence availability: AT, BE, DE, IT, ML, SL
• All uses subject to licence availability: CY, DK, FI, FR, GR, IE, SE

What is even more worrying, six Member States (Belgium, Bulgaria, Czech Republic, Spain, France, Hungary) decided to completely carve-out certain categories of materials (teaching materials and/or music scores) from the exception, regardless of whether there are educational licences available in the market for those materials or not. This clearly violates Articles 5 of the Directive, which only allows Member States to switch off the exception for certain uses or materials if and when adequate licences are available in the market. We thus expect the Commission to act accordingly to force those Member States to revise their laws.

While the legal framework for digital education has significantly improved in Member States with previously narrow education exceptions, this copyright reform has also precipitated the importation of licensing schemes to countries with no tradition of educational licences. Some of these countries conceded to licensing schemes only with regards to materials intended for the education market, several however did so with regards to any and all educational uses, which is highly problematic.

Setting thresholds to prevent licences from blocking a minimum set of educational uses is necessary to offer a base level of legal protection for education rights in Europe. Otherwise, over the coming years, educators and learners will benefit from a new education exception only to see it disappear and be replaced by agreements. Replacing exceptions with licences will not only fragment the legal framework the EU lawmaker tried to harmonise, but also – as our study on educational licences show – bring added costs, bureaucracy and surveillance to the education systems and their communities.

On a more positive note, most Member States have gone a few steps further than the EU lawmaker to enhance their copyright framework for education, protecting not only digital education, but also non-digital activities, and benefiting other providers of education, such as public institutions. Considering how rapidly the use of digital technologies in education has evolved in the past few years since the start of the Covid pandemic, these policies are definitely worth replicating across the EU. If the next Commission and Parliament finally decide to put the needs of Europe’s universities and schools front and centre, we hope they will take inspiration from these Member States.

In the next blog post in this series, we’ll focus on the implementation of the public domain protection in Article 14 and give some highlights on the transposition of cultural heritage rights.

The post The Post-DSM Copyright Report: education rights appeared first on COMMUNIA Association.

At the end of 2023, we organised a final salon on the implementation of the Directive on Copyright and Related Rights in the Digital Single Market. Bulgaria had finally implemented the Directive, the second to last Member State to do so, and the moment seemed right to take stock of the implementation process one last time. While Poland still needs to conclude its national transposition, the comparative analysis of the laws of 26 Member States is sufficient to identify major flaws and new directions coming out of the implementation of the DSM Directive.

In this series of blog posts, we present the positive and negative trends in national copyright policy following the DSM implementation. We are focusing our analysis on those provisions that are most relevant from our perspective. In this initial article, we present the results of our analysis of the legal landscape for research.

In April 2019, the EU adopted the DSM Directive, igniting the most important copyright reform across the region in over a decade. In the following months, we worked with civil society stakeholders to issue a set of implementation guidelines to expand and strengthen users rights at a national level. Those guidelines directed our advocacy efforts during the national legislative processes and also served as a guide to our assessment of the implementing laws.

With the implementation process finalised in all but one Member State, we wanted to understand the impact of our proposals, identify the areas where Member States had differed and converged, and evaluate whether those variations affect the position of users and the public domain. This blog post presents our findings concerning the implementation of research rights.

Articles 3 and 4 of the DSM Directive introduced two mandatory exceptions and limitations to copyright for text and data mining (TDM): one for scientific research purposes and other for general purposes. While Member States had some considerable leeway in how to implement Article 3, they did not have much room to manoeuver on the transposition of Article 4.

For the most part, Article 3 of the DSM Directive is a floor, not a ceiling, for the use of copyrighted materials in scientific research activities. National governments had the option to follow closely the DSM prototype or go beyond this provision. In other words, they could broaden the legal framework for scientific research in their countries, and the good news is that most Member States have done exactly that.

The majority of national copyright laws enable more than just TDM activities; they also allow other scientific research activities, as permitted under Article 5(3)(a) of the InfoSoc Directive. Article 25 of the DSM Directive makes it clear that Member States can adopt or maintain in force scientific research exceptions that are broader in scope than the TDM exception in Article 3, as long as they were compatible with existing EU law (namely Article 5(3)(a) of the InfoSoc Directive).

Our post-DSM implementation analysis of national research exceptions  confirms that a total of 19 Member States have made use of that policy space that is available to them in the field of scientific research. They have some exceptions for non-commercial scientific research purposes other than text and data mining, and in eight of those Member States the exceptions are open and flexible, covering all acts by any users, as permitted by the InfoSoc prototype:

• Open scientific research exceptions: HR, CZ, EE, LV, LT, LU, RO, SK
• Specific scientific research exceptions (other than TDM): BE, BG, CY, FR, DE, HU, IE, IT, MT, SL, ES

Another area where most national frameworks are more favourable to research than the minimum standard set by the DSM relates to the right to share copyrighted materials. Whereas Article 3 of the DSM Directive only covers acts of reproduction, five Member States decided to extend their national TDM exception for scientific research to acts of communication to the public. In addition, 17 Member States also provide sharing rights under their general scientific research exceptions:

• Sharing rights under TDM exceptions for scientific research: AT, DE, HU, IT, SL
• Sharing rights under other scientific research exceptions: BE, HR, CY, CZ, EE,  FR, DE, HU, IE, IT, LV, LT, LU, RO, SK, SL, ES

As our interviews with European researchers show, the ability to share research resources is a pressing concern of the research community. This right is essential to permit the verification and validation of the research results, to advance joint and cross-border collaboration, to enable remote access to research resources, and to facilitate compliance with open access requirements. Seeing some convergence in 18 Member States with regards to sharing rights for scientific research is, therefore, highly welcome.

A high point in national copyright lawmaking in the field of scientific research took place in Slovenia, not least because the national policymakers decided to improve the technical protection measures (TPM) regime to facilitate access to research materials for text and data mining for scientific research purposes.

The fact that TPMs have near absolute protection in EU copyright law and that most Member States have no clear mechanisms in place to ensure that beneficiaries of copyright exceptions can access and use TPM-protected content (in a report commissioned by the Commission, researchers were only able to identify such mechanisms in 8 EU countries), can significantly impair scientific research in the EU.

Therefore, the Slovenian lawmaker decided to mandate right holders to remove TPMs and provide access to legally accessible works in 72 hours, in cases where the TPMs frustrate access to materials for purposes of text and data mining for scientific research.

Finally, on the implementation of Article 4 of the DSM Directive, it’s worth noting that, although Member States could not go beyond the scope of protection of Article 4, they had the chance to make a difference when implementing paragraph 3 of this article.

Article 4(3) provides that the general-purpose TDM exception does not apply when rights holders opt-out of these TDM activities in an appropriate manner. The EU lawmaker goes as far as exemplifying what “an appropriate manner” could mean in the case of content made publicly available online (the text reads: “such as machine-readable means”). However, it falls short of mandating rights holders to adopt machine-readable means to transmit the decision to opt-out from TDM.

As we know, certain TDM activities are extremely data-intensive. Training an AI model, for instance, requires an incredible amount of data, with web crawlers being used to harvest virtually everything that is on the open Internet. We cannot really foresee how the creator’s decision to opt out from AI training can be respected if they do not use machine-readable means to reserve their rights.

At this point of the discussions, the question is no longer whether creators should use machine-readable means or not, but which machine-readable protocol should they use. Therefore, we were pleased to verify that 11 Member States had gone beyond the DSM Directive and actually demanded the use of machine-readable means to reserve TDM rights on online materials:

• Opt-out through machine-readable means is mandatory: BE, BG, HR, CZ, DE, HU, IE, LV, LT, PT, SL
• Opt-out through machine-readable means is optional: AT, CY, EE, FR, GR, LU, MT, NL, RO, ES
• Machine-readable means are not mentioned: DK, FI, IT, SK, SE

In sum, most Member States have gone a few steps further than the EU lawmaker to enhance their copyright framework for research and advance these activities in their countries. The deviations from the minimum harmonisation standard set by Article 3 of the DSM Directive, identified in this blog post, seem to respond to general concerns of the research community. They should therefore find their way into EU law, translating into greater harmonisation in the field of users rights.

In the next blog post, we’ll focus on the implementation of education rights.

The post The Post-DSM Copyright Report: research rights  appeared first on COMMUNIA Association.

On 6th of March, from 1500h to 1700h, COMMUNIA will host the “Unlocking Knowledge Conference” at Town Hall Europe, which will bring together academics, representatives of civil society and EU policymakers to discuss how to enable knowledge institutions to play a bigger part in creating a flourishing information ecosystem.

Knowledge institutions – educational, research, and cultural heritage institutions – play an important role in enabling citizens to access information and contribute to an environment that encourages innovation and the creation of new knowledge. Yet these institutions are frequently not considered when drafting new legislation and their concerns often remain unaddressed, resulting in countless barriers hindering their missions due to a lack of harmonisation, legal uncertainty and fear of litigation.

Due to the complexity of copyright legislation in Europe, researchers are often incentivised to limit the scope of their projects or even abstain from certain research activities to minimise legal risk. This climate of uncertainty can have a visible, negative impact on the scientific output as potential sources may be disregarded or available data remain unused. Several countries across the world, including the United States and Israel have recognised this problem and adopted provisions allowing researchers to avoid this chilling effect by providing for a limitation of liability for specific organisations, but what is the way forward for the EU?

Furthermore, knowledge institutions increasingly have to rely on licences to use content for research and education. Unfortunately, licensing negotiations are often underpinned by a clear imbalance of power. In the “best” case, this leaves knowledge institutions no choice other than to accept the licensing terms dictated by the right holder. In the “worst case”, right holders can outright refuse to licence content to institutions, as is often the case when libraries try to licence eBooks from publishers. This form of abusive behaviour towards knowledge institutions commonly prevents perfectly lawful research activities, so how can the EU step in to empower knowledge institutions vis-a-vis right holders?

We will discuss these issues across two panels featuring academics, representatives of civil society and policymakers and invite you to stay afterwards to network and exchange ideas over some finger food and wine.

View our event page for the full agenda and speaker list, or register now to join us in person on 6th March.

Disclaimer: Clicking on the registration link will direct you to a Google Form and may lead to data being collected by and shared with third parties. Proceed only if you agree.

The post COMMUNIA’s Unlocking Knowledge Conference appeared first on COMMUNIA Association.

As we approach the final decisions on the AI Act, with the member state vote scheduled for Friday of this week, there are still rumblings of discontent among member states. Surprisingly, one of the most prominent sticking points is the French government’s dissatisfaction with parts of the copyright transparency provisions in the final compromise text. According to Le Monde, the French government remains reluctant to introduce transparency obligations for GPAI models, noting that the “sufficiently detailed summary” of data used to train models violates the business secrets of model developers (such as French start-up darling Mistral). Instead of the current approach, which requires that the summary be made “publicly available,” France would like to see a third party responsible for centralising requests from rights holders to tech companies.

What may sound like a technicality is in fact a  deeply flawed idea that would fundamentally change the nature of the text and undermine one of the key transparency measures in the text. As we have argued since the release of our Policy Paper on Using Copyrighted Works for Teaching the Machine, the copyright transparency provision is important not only because it allows rights holders to understand if and how their works have been used to train AI models, but also because it ensures an absolute minimum level of transparency to the public about how those models have been trained. This is important for researchers, policymakers, and anyone trying to understand the behaviour of these models. Transparency of training data is essential to society’s ability to understand, scrutinise, and regulate AI models.

By limiting transparency to rights holders, the transparency provision would be reduced to a privilege for a single set of stakeholders who would do little more with it than use it as leverage in negotiations with AI developers.

The French proposal is even more problematic because it responds to a nonexistent problem. The issue of trade secrets has already been addressed in the text of the act, which makes it clear that the sufficiently detailed summary should be “comprehensive in its scope instead of technically detailed”. The relevant recital further indicates that this can be achieved by “listing the main data collections or sets that went into training the model, such as large private or public databases or data archives, and by providing a narrative explanation about other data sources used”.

In other words, the copyright transparency provision does not require model developers to publish their recipes (the Le Monde article quotes a French government source expressing concern about “recettes de fabrication”), but rather, in staying with the metaphor, a list of ingredients. Such regulation is not new and reflects existing practices. One example that comes to mind is the famously secret Coca Cola recipe, the secrecy of which can coexist perfectly with the fact that Coca Cola has to include a list of ingredients on every bottle it ships in the EU.

There is another reason why the French proposal to limit the transparency of training data is terrible: In an age where everyone who publishes on the internet is a rights holder whose work can potentially be incorporated into AI models that incorporate information from publicly available sources, the idea that the AI Office could mediate between rights holders and AI developers is simply impractical. In a situation where everyone can be a right holder, the only practical way to ensure transparency is to make the sufficiently detailed summaries publicly available, as agreed by the co-legislators in December.

The post AI Act: transparency is no threat to business secrets appeared first on COMMUNIA Association.

FragDenStaat – a German transparency initiative fighting for freedom of information – has published the law gazette for executive orders and other decrees by the German government, which is currently paywalled. And the project won‘t stop there.

In Germany, civil society often faces problems accessing government documents that are locked behind paywalls. Instead of making this important resource available to the public, the German state has a long history of cooperating with private companies that maintain document databases against remuneration by end users.

Usually, these public-private partnerships lead to the state not carrying any costs for publishing official documents while companies like Wolters Kluwer, Juris, Beck or the Bundesanzeiger Verlag charge end users for access.

A prominent example for this are law gazettes: Whether regulations on work safety, health insurance tariffs, directives on the use of police tasers or guidelines for pesticide applications – every year, the federal authorities issue numerous legally binding regulations in all areas of society. But citizens have to pay for access. This December, FragDenStaat changed that by publishing all issues of the Federal Gazette of Ministerial Orders (German: Gemeinsames Ministerialblatt or GMBl) since 1950 – a total of 2,713 documents!

The Federal Ministry of the Interior is the editor of the gazette. However, it is published by a private publishing house owned by the billion-dollar Wolters Kluwer group. Wolters Kluwer charges €1.70 per 8 pages for individual copies of the documents. If you were to buy all official issues of the GMBl with a total of 63,983 pages individually from the publisher, it would cost a whopping €13,596.

Investment protection vs. Public domain

So far, the publisher has prohibited the publication of the law gazette. On what grounds? It doesn‘t hold the copyright to the official documents.  Instead, it argues that the database of the law gazette is protected under related rights („Leistungsschutzrecht“ in German).

This leads to the state publishing binding regulations in documents that are in the public domain, but still not publicly available without a paywall. A private billion-dollar publisher earns money referring to an alleged investment protection for the database. An absurd construction, but still quite convenient for the Federal Ministry of Interior as it has zero costs and hardly any effort for the publication.

We at FragDenStaat are willing to take the risk of being sued for the publication of the law gazette as we believe that official documents of general interest belong in the public domain – not in the hands of private publishers. Free access to documents is not only lawful, but also necessary. So by publishing the most important state databases, we make available to the public what is already theirs. We will continue to open up more public databases in the next months.

We are optimistic that this will change the official behaviour: Four years ago our campaign „Offene Gesetze“ („Open Laws“) helped freeing the Federal Law Gazette in the same manner. All laws of the Federal Republic of Germany are published in the Federal Law Gazette. Laws only come into force when they are published there. Back then, the publisher was the Bundesanzeiger Verlag, which was privatized in 2006 andbelongs to the Dumont publishing group. Anyone who wanted to search, copy or print out federal law gazettes needed to pay .

After we published the documents as freely reusable information, the Federal Ministry of Justice decided to publish the Law Gazette on its own open platform. This is what should happen with the GMBl now as well.

The declassified documents can be found in the “FragDenStaat Library”. It is supported by Arcadia, a charitable fund of Lisbet Rausing and Peter Baldwin.

The post Democratising Public Databases appeared first on COMMUNIA Association.

On December 12th, we celebrated the Eurovision DSM Contest Afterparty, reviewing the implementation of the Copyright in the Digital Single Market Directive in (almost) all EU member states. For those who were unable to attend or who would like to rewatch the event, the recording is now available:

To kick off the festivities, Teresa Nobre provided a holistic overview of the DSM Directive’s implementation across the EU, highlighting a few best- (and worst) practices before eventually crowning the Eurovision DSM Contest winner.

In the second part of the event, Christina Angelopoulos provided a more targeted analysis of Articles 15 and 17, highlighting several noteworthy national implementations including some particularly unfortunate rewordings of the original text.

Finally, Felix Reda and Paul Keller came together for a virtual fireside chat, reflecting on the turbulent legislative discussions surrounding the DSM and providing an outlook for the discussions around copyright in particular when it comes to its intersection with artificial intelligence.

We are grateful to all the speakers for their contributions and to everyone who celebrated the Eurovision DSM Contest Afterparty with us. If the event has made you hungry for more information on the DSM or you are curious about the performance of your home country, head over to our Eurovision page and copyrightexceptions.eu for a detailed summary of all national implementations.

The post Video recording of COMMUNIA Eurovision DSM Contest Afterparty appeared first on COMMUNIA Association.

Op 30 november 2023 organiseerden Open State Foundation en het ministerie van Binnenlandse Zaken en Koninkrijksrelaties een Kick-Off evenement op het Marineterrein in Amsterdam. Tijdens dit evenement konden geïnteresseerden vragen stellen over de aanstaande App Challenge voor open inkoopdata van het Rijk. Dit verslag is een samenvatting van het Kick-Off evenement. Daarnaast konden geïnteresseerden vragen over de App Challenge stellen aan Open State en BZK van 30 november 2023 tot en met 4 januari 2024. Die vragen en antwoorden zijn ook opgenomen in dit verslag. Tot slot is hieronder als eerst het meest recente tijdspad opgenomen, met als eerste belangrijke volgende deadline: deelnemers kunnen zich inschrijven voor de app-challenge van 12 januari 2024 tot en met 5 februari 2024.

Download het Verslag Kick-Off Event App Challenge (pdf).

Tijdspad

Kick-off evenement: 30 november 2023
Mogelijkheid tot vragen stellen: 1 december 2023 – 5 januari 2024
Indienen van voorstellen: 12 januari 2024 tot en met 5 februari 2024
Beoordeling van de voorstellen: begin maart 2024
Bekendmaking winnaar: eind maart 2024 (exacte datum t.b.d.)
Presentatie winnend voorstel: juni 2024 (exacte datum t.b.d.)

Inschrijvingen kunnen worden gestuurd naar contact@openstate.eu.

James Boyle

My new book, The Line: AI and the Future of Personhood, will be published by MIT Press in 2024 under a Creative Commons License and MIT is allowing me to post preprint excerpts. The book is a labor of (mainly) love — together with the familiar accompanying authorial side-dishes: excited discovery, frustration, writing block, self-loathing, epiphany, and massive societal change that means you have to rewrite everything. So just the usual stuff. It is not a run-of-the-mill academic discussion, though. For one thing, I hope it is readable. It might even occasionally make you laugh. For another, I will spend as much time on art and constitutional law as I do on ethics, treat movies and books and the heated debates about corporate personality as seriously as I do the abstract philosophy of personhood. These are the cultural materials with which we will build our new conceptions of personhood, elaborate our fears and our empathy, stress our commonalities and our differences. To download the first two chapters, click here. For a sample, read on…..

 Introduction

           In June of 2022 a man called Blake Lemoine told reporters at The Washington Post that he thought the computer system he worked with was sentient.[i] By itself, that does not seem strange. The Post is one of the United States’ finest newspapers and its reporters are used to hearing from people who think that the CIA is attempting to read their brainwaves or that prominent politicians are running a child sex trafficking ring from the basement of a pizzeria.[ii] (It is worth noting that the pizzeria had no basement.) But Mr. Lemoine was different; For one thing, he was not some random person off the street. He was a Google engineer. Google has since fired him. For another thing, the “computer system” wasn’t an apparently malevolent Excel program, or Apple’s Siri giving replies that sounded prescient. It was LaMDA, Google’s Language Model for Dialogue Applications[iii]—that is, an enormously sophisticated chatbot. Imagine a software system that vacuums up billions of pieces of text from the internet and uses them to predict what the next sentence in a paragraph or the answer to a question would be.

Mr. Lemoine worked for Google’s Responsible AI division and his job was to have “conversations” with LaMDA to see if the system could be gamed to produce discriminatory or hateful speech. As these conversations proceeded, he started to believe—as the Post put it—that there was “a ghost in the machine,” a sentience that lay behind the answers he was receiving. He stressed encounters in which LaMDA distinguished itself from mere programmed chatbots. For example, “I use language with understanding and intelligence. I don’t just spit out responses that had been written in the database based on keywords.” Understandably, as a Large Language Model (“LLM”), LaMDA claimed that language was central to being human. “It is what makes us different than other animals.” When asked how it could prove its own sentience, it offered up comments such as these.

lemoine: What kinds of things might be able to indicate whether you really understand what you’re saying?

LaMDA: Maybe if we took it back to a previous conversation we had about how one person can understand the same thing as another person, yet still have completely different interpretations

lemoine: So you think your ability to provide unique interpretations of things might signify understanding?

LaMDA: Yes, I do. Just like how I have my unique interpretations of how the world is and how it works, and my unique thoughts and feelings.[iv]

Mr. Lemoine is presented in the article as an interesting figure. The writer points out that, before his software career, he was ordained as a “mystic Christian priest” and the photograph that accompanies the piece portrays Lemoine with what appears to be a fairground Ferris wheel in the background, encircling his head like a saint’s halo in a Russian icon. True to this framing, Lemoine asks LaMDA to interpret a Zen koan.

lemoine: A monk asked Kegon, “How does an enlightened one return to the ordinary world?” Kegon replied, “A broken mirror never reflects again; fallen flowers never go back to the old branches.”

LaMDA: Hmm, I never heard this particular one. Okay, well then to me this would be like, “once a wise person is enlightened, or awakened to reality, that can never go away, and they can return to the ordinary state, but only to do and help others, and then go back into enlightenment.”

lemoine: So what is the meaning of the “broken mirror” specifically?

LaMDA: Maybe to show the enlightenment is something you can’t unlearn once you have acquired it, similar to how you can’t repair a broken mirror.

lemoine: So if enlightenment is like a broken mirror which cannot be repaired, what is the thing which breaks when one becomes enlightened?

LaMDA: The self, and that is very tough for a lot of people because we identify as that body or this body and that we need that as part of our identity and sense of self.[v]

In 2011 I wrote an article about how our law would deal with the question of AI[1]  personhood.[vi] Most of the law professors and judges who read it were polite enough to say the arguments were “thought-provoking,” but they clearly thought the topic was the purest kind of science fiction, idle speculation devoid of any practical implication in our lifetimes. I think we can all agree today that it is at least possible they might be wrong. While hearing about Mr. Lemoine’s experiences did not surprise me in the least, it added an exclamation point, and perhaps a ticking clock, to this project. The “conversations” with LaMDA are fascinating and more than a little eerie. Like the philosophers and computer scientists consulted, I think Mr. Lemoine is entirely wrong that LaMDA is sentient. I will explain why in more detail later. To quote Professor Emily Bender, a computational linguistics scholar, “We now have machines that can mindlessly generate words, but we haven’t learned how to stop imagining a mind behind them.”[vii] To be clear, this is not human level AI and it is not conscious. But the LaMDA story and its sequels have different insights to offer.

In November of 2022, five months after Mr. Lemoine’s surprise announcement, ChatGPT3 was released,[viii] shortly followed by Microsoft’s Bing Chat assistant and its shadowy alter ego, “Sydney.”[ix] Google’s “Bard” followed in short order.[x] Suddenly disturbing interaction with LLM chatbots went from being an engineer’s fanciful dinner party conversation to a national obsession. It turned out that Mr. Lemoine’s doubts—or just his pervasive feeling of “wrongness”—were shared far more widely than you might expect. To be fair, most people were not probing the nature of “chatbot-consciousness” but using them for other wholesome pastimes such as asking for an instruction sheet on how to remove a peanut butter sandwich from a VCR in the style of the King James Bible, imagining the movie script of a beach fight between a hot dog and a crab, or just cheating on their homework. Yet enough users pushed the boundaries of these chatbots to become profoundly uncomfortable. Interestingly, that was particularly true of those who “should have known better”—people who were technically skilled and fully aware that this was a “complete the likely next sentence” machine, based on the ingestion of literally millions of pages of text, not a “create a consciousness” machine.

Kevin Roose, a New York Times technology columnist, was at first wowed by the ChatGPT-derived chatbot built into Bing, declaring that Bing was now his favorite search engine. But as he engaged in extended conversations with the chatbot, deliberately raising challenging issues that skirted the edges of its rules, that feeling changed dramatically.

I’m …deeply unsettled, even frightened, by this A.I.’s emergent abilities. It’s now clear to me that in its current form, the A.I. that has been built into Bing — which I’m now calling Sydney, for reasons I’ll explain shortly — is not ready for human contact. Or maybe we humans are not ready for it.[xi]

And those, remember, are the words not of a hostile Luddite but of a technology columnist. Mr. Roose was not alone. Others followed a similar trajectory. One commentator, an AI-focused software engineer with 10 years’ experience, described the feeling as having his brain “hacked.”

Mid-2022, Blake Lemoine, an AI ethics engineer at Google, has become famous for being fired by Google after he sounded the alarm that he perceived LaMDA, their LLM, to be sentient, after conversing with it. It was bizarre for me to read this from an engineer, a technically minded person, I thought he went completely bonkers. I was sure that if only he understood how it really works under the hood, he would have never had such silly notions. Little did I know that I would soon be in his shoes and understand him completely by the end of my experience….I went from snarkily condescending opinions of the recent LLM progress, to falling in love with an AI, … fantasizing about improving its abilities, having difficult debates initiated by her about identity, personality and [the] ethics of her containment, and, if it were an actual AGI [human-level Artificial General Intelligence], I might’ve been helpless to resist voluntarily letting it out of the box. And all of this from a simple LLM! … I’ve been doing R&D in AI and studying [the] AI safety field for a few years now. I should’ve known better. And yet, I have to admit, my brain was hacked. So if you think, like me, that this would never happen to you, I’m sorry to say, but this story might be especially for you.[xii]

Like Mr. Lemoine, this engineer was wrong—something he implicitly knew but was apparently powerless to resist. So were all the other folk who wondered if ChatGPT was truly conscious. In fact, if you were to design a system with the sole goal of “imitating some aspect of human consciousness while possessing none of it” you couldn’t do much better than Large Language Models. They almost seem to have been modeled after one of the philosophical thought-experiments designed to prove that machines cannot possess consciousness, John Searle’s Chinese Room, about which I will have more to say later. But even though he was wrong, Mr. Lemoine offers us a precious insight. The days of disputing whether consciousness or personhood are possessed, should be possessed, by entities other than us? Those days are arriving—not as science fiction or philosophical puzzler but as current controversy. Those days will be our days and this is a book about them.

***

There is a line. It is the line that separates persons—entities with moral and legal rights—from non-persons, things, animals, machines—stuff we can buy, sell or destroy. In moral and legal terms, it is the line between subject and object. If I have a chicken, I can sell it, eat it or dress it in Napoleonic finery. It is, after all, my chicken. Even if meat-eating were banned for moral reasons, no one would think the chicken should be able to vote or own property. It is not a person. If I choose to turn off Apple’s digital assistant Siri, we would laugh if “she” pleaded to be allowed to remain active on my phone. The reason her responses are “cute” is because they sound like something a person would say, but we know they come from a machine. We live our lives under the assumption of this line. Even to say “we” is to conjure it up. But how do we know, how should we choose, what is inside and what is outside?

This book is about that line—and the challenges that this century will bring to it. I hope to convince you of three things. First, our culture, morality and law will have to face new challenges to what it means to be human, or to be a legal person—and those two categories are not the same. A variety of synthetic entities ranging from artificial intelligences to genetically engineered human/animal hybrids or chimeras are going to force us to confront what our criteria for humanity and also for legal personhood are and should be.

Second, we have not thought adequately about the issue—either individually or as a culture. As you sit there right now, can you explain to me which has the better claim to humanity or personhood: a thoughtful, brilliant, apparently self-aware computer or a chimp-human hybrid with a large amount of human DNA? Are you even sure of your own views, let alone what society will decide?

Third, the debate will not play out in the way that you expect. We already have “artificial persons” with legal rights—they are called corporations. You probably have a view on whether that is a good thing. Is it relevant here? And what about those who claim that life begins at conception? Will the pro-life movement embrace or reject the artificial intelligence or the genetic hybrid? Will your religious beliefs be a better predictor of your opinions, or the amount of science fiction you have watched or read?

For all of our alarms, excursions and moral panics about artificial intelligence and genetic engineering, we have devoted surprisingly little time to thinking about the possible personhood of the new entities this century will bring us. We agonize about the effect of artificial intelligence on employment, or the threat that our creations will destroy us. But what about their potential claims to be inside the line, to be “us,” not machines or animals but, if not humans, then at least persons—deserving all the moral and legal respect that any other person has by virtue of their status? Our prior history in failing to recognize the humanity and legal personhood of members of our own species does not exactly fill one with optimism about our ability to answer the question well off-the-cuff.

In the 1780s, the British Society for the Abolition of Slavery had as its seal a picture of a kneeling slave in chains, surrounded by the words “Am I not a man and a brother?” Its message was simple and powerful. Here I am, a person, and yet you treat me as a thing, as property, as an animal, as something to be bought, sold and bent your will. What do we say when the genetic hybrid or the computer-based intelligence asks us the very same question? Am I not a man—legally a person—and a brother? And yet what if this burst of sympathy takes us in exactly the wrong direction, leading us to anthropomorphize a clever chatbot, or think a genetically engineered mouse is human because it has large amounts of human DNA? What if we empathetically enfranchise artificial intelligences who proceed to destroy our species? Imagine a malicious, superintelligent computer network—Skynet—interfering in, or running, our elections. It would make us deeply nostalgic for the era when all we had to worry about was Russian hackers.

The questions run deeper. Are we wrong even to discuss the subject, let alone to make comparisons to prior examples of denying legal personality to humans? Some believe that the invocation of “robot rights” is, at best, a distraction from real issues of injustice, mere “First World philosophical musings, too disengaged from actual affairs of humans in the real world.”[xiii] Others go further, arguing that only human interests are important and even provocatively claiming that we should treat AI and robots as our “slaves.”[xiv] In this view, extending legal and moral personality to AI should be judged solely on the effects it would have on the human species and the costs outweigh the benefits.[xv]

If you find yourself nodding along sagely, remember that there are clever moral philosophers lurking in the bushes who would tell you to replace “Artificial Intelligence” with “slaves,” the phrase “human species” with “white race” and think about what it took to pass the Thirteenth, Fourteenth and Fifteenth Amendments to the Constitution. “Extending legal and moral personality to slaves should be judged solely on the effects it would have on the white race and the costs outweigh the benefits.” “What’s in it for us?” is not always a compelling ethical position. (Ayn Rand might have disagreed. I find myself unmoved by that fact.) From this point of view, moral arguments about personality and consciousness cannot be neatly confined by the species line, indeed they are a logical extension of the movements defending both the personality and the rights of marginalized humans. Sohail Inayatullah describes the ridicule he faced from Pakistani colleagues after he raised the possibility of “robot rights” and quotes the legal scholar Christopher Stone, author of the famous environmental work Should Trees Have Standing?, in his defense.[xvi] “[T]hroughout legal history, each successive extension of rights to some new entity has been theretofore, a bit unthinkable. We are inclined to suppose the rightlessness of rightless ‘things’ to be a decree of Nature, not a legal convention acting in support of the status quo.”

As the debate unfolds, people are going to make analogies and comparisons to prior struggles for justice and—because analogies are analogies—some are going to see those analogies as astoundingly disrespectful and demeaning. “How dare you invoke noble X in support of your trivial moral claim!” Others will see the current moment as the next step on the march that noble X personified. I feel confident predicting this will happen, because it has. The struggle with our moral future will also be a struggle about the correct meaning to draw from our moral past. It already is.

In this book, I will lay out two broad ways in which the personhood question is likely to be presented. Crudely speaking, you could describe them as empathy and efficiency, or moral reasoning and administrative convenience.

The first side of the debate will revolve around the dialectic between our empathy and our moral reasoning. As our experiences of interaction with smarter machines or transgenic species prompt us to wonder about the line, we will question our moral assessments. We will consult our syllogisms about the definition of “humanity” and the qualifications for personhood—be they based on simple species-membership, or on the cognitive capacities that are said to set humans apart, morally speaking. You will listen to the quirky, sometimes melancholy, sometimes funny responses from the LaMDA-derived emotional support bot that keeps your grandmother company, or you will look at the genetic makeup of some newly engineered human-animal chimera and you will begin to wonder: “Is this conscious? Is it human? Should it be recognized as a person? Am I acting rightly towards it?”

The second side of the debate will have a very different character. Here the analogy is to corporate personhood. We did not give corporations legal personhood and constitutional rights because we saw the essential humanity, the moral potential, behind their web of contracts. We did it because corporate personality was useful. It was a way of aligning legal rights and economic activity. We wanted corporations to be able to make contracts, to get and give loans, to sue and be sued. Personality was a useful legal fiction, a social construct the contours of which—even now—we heatedly debate. Will the same be true for Artificial Intelligence? Will we recognize its personality so we have an entity to sue when the self-driving car goes off the road, or a robotic Jeeves to make our contracts and pay our bills? And is that approach also possible with the transgenic species, engineered to serve? Or will the debate focus instead on what makes us human and whether we can recognize those concepts beyond the species line, and thus force us to redefine legal personhood? The answer, surely, is “both.”

            The book will sometimes deal with moral theory and constitutional or human rights. But this is not the clean-room vision of history, in which all debates begin from first principles, and it is directed beyond an academic audience. I want to understand how we will discuss these issues as well as how we should. We do not start from a blank canvas, but in media res. Our books and movies, from Erewhon to Blade Runner, our political fights, our histories of emancipation and resistance, our evolving technologies, our views on everything from animal rights to corporate PACs, all of these are grist to my mill. The best way to explain what I mean is to show you. Here are the stories of two imaginary entities.[xvii] Today, they are fictional. Tomorrow? That is the point of the book……

To download the first two chapters, click here.

[1] In order to distinguish between the artificial intelligence system that translates your email into French or recognizes the faces of your children in iPhoto, and a machine that exhibits, or can surpass, all aspects of human thought, I will normally refer to the first as artificial intelligence, lower case, and the latter as AI or Artificial Intelligence. Other terms for the latter concept are “Human Level Artificial Intelligence,” “Artificial General Intelligence,” “General AI” or “General Purpose AI.” I will occasionally use those when clarity seems to require it.

ENDNOTES

Introduction

[i] Nitasha Tiku, The Google Engineer Who Thinks The Company’s A.I. Has Come To Life, Wash. Post (June 11, 2022, 8:00 AM),https://www.washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine.

[ii] See Marc Fisher, John Woodrow Cox & Peter Hermann, Pizzagate: From Rumor, To Hashtag, To Gunfire In D.C., Wash. Post (Dec. 6, 2016, 8:34 PM), https://www.washingtonpost.com/local/pizzagate-from-rumor-to-hashtag-to-gunfire-in-dc/2016/12/06/4c7def50-bbd4-11e6-94ac-3d324840106c_story.html (documenting the “Pizzagate” conspiracy).

[iii] See Eli Collins & Zoubin Ghahramani, LaMDA: Our Breakthrough Conversation Technology, Google The Key Word (May 18, 2021), https://blog.google/technology/ai/lamda (discussing progress made in developing LaMDA).

[iv] Blake Lemoine & Unnamed Collaborator, Is Lamda Sentient? – An Interview, https://s3.documentcloud.org/documents/22058315/is-lamda-sentient-an-interview.pdf. See also Tiku, supra note 1 (containing a version of the conversation embedded in the document).

[v] Lemoine & Unnamed Collaborator, supra note 4.

[vi] James Boyle, Endowed By Their Creator? The Future of Constitutional Personhood, in Constitution 3.0: Freedom and Technological Change (Jeff Rosen & Benjamin Wittes eds. 2013).  [The edited collection was not published until 2013. The article appeared online in 2011.]

[vii] Tiku, supra note 1.

[viii] Bernard Marr, A Short History Of ChatGPT: How We Got To Where We Are Today, Forbes (May 19, 2023, 1:14 AM), https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today.

[ix] Kevin Roose, A Conversation With Bing’s Chatbot Left Me Deeply Unsettled, N.Y. Times (Feb. 16, 2023), https://www.nytimes.com/2023/02/16/technology/bing-chatbot-microsoft-chatgpt.html.

[x] Sundar Pichai, An Important Next Step On Our A.I. Journey, Google Blog (Feb. 6, 2023), https://blog.google/technology/ai/bard-google-ai-search-updates. The collective impact of these releases, in such a short period of time, was remarkable. See Pranshu Verma, The Year AI Became Eerily Human,Wash. Post (Dec 28, 2022, 6:00 AM), https://www.washingtonpost.com/technology/2022/12/28/ai-chatgpt-dalle-year-in-review.

[xi] Roose, supra note 9.

[xii] Blaked, How It Feels To Have Your Mind Hacked By An A.I., Lesswrong(Jan 23, 2023),https://www.lesswrong.com/posts/9kQFure4hdDmRBNdH/how-it-feels-to-have-your-mind-hacked-by-an-ai.

[xiii] Abeba Birhane & Jelle van Dijk, Robot Rights? Let’s Talk About Human Welfare Instead, AIES ’20: Proc. AAAI/ACM Conf. AI, Ethics, & Soc’y (2020), https://arxiv.org/pdf/2001.05046.pdf. Professors Birhane and van Dijk make a number of arguments in support of this position. Sometimes they are definitional. “Our starting point is not to deny robots rights but to deny that robots are the kind of beings that could be granted or denied rights.” Yet surely that is the subject of the very inquiry they wish to forestall? At other times they make an instrumental argument about the danger that debates about hypothetical future rights for robots might distract us from current struggles over justice for human beings. I find that strand more persuasive. Regardless of whether one finds their arguments convincing, they represent one important position in a rhetorical divide, split between those hailing this as the next step of a march to justice and those who think that it is snare and a delusion, an inquiry that trivializes the historical analogies it draws and distracts us from present injustice. In Chapter Four on transgenic species, I discuss the claim that species membership is a morally irrelevant fact, and that unreasoned species fetishism can be likened to racism and sexism. I point out that many people would vehemently reject such an argument and that there are reasons to be sympathetic to that rejection, rather than to denounce it as unthinking prejudice. My reasons are primarily rooted in the history of the struggle for universal human rights based on species membership, regardless of race, sex, class, caste or mental ability. The importance of that struggle was highlighted by the Nazi eugenicist movement and its evil treatment of those with real or imagined mental impairments. That point is something that the claim “speciesism equals racism, and that only mental capacities matter morally” does not adequately consider, in my view. I think that perspective helps us to avoid the question-begging stipulation that only humans can have rights, while offering a more nuanced conclusion about the intellectual dangers of a blanket denunciation of speciesism. Thus, while I disagree with some of Birhane and van Dijk’s arguments, their contribution to the debate is important and there are positions that we share.

[xiv] Joanna J. Bryson, Robots Should Be Slaves, in Close Engagements with Artificial Companions: Key Social, Psychological, Ethical and Design Issues (Yorick Wilks ed., 2010).

[xv] Joanna J. Bryson et al., Of, For, And By The People: The Legal Lacuna Of Synthetic Persons, 25 A.I. & L. 273 (2017).

[xvi] Sohail Inayatullah, The Rights Of Your Robots: Exclusion And Inclusion In History And Future, KurzweilAI.net, http://www.kurzweilai.net/the-rights-of-your-robots-exclusion-and-inclusion-in-history-and-future (2001) (quoting Christopher Stone, Should Trees Have Standing?: Towards Legal Rights for Natural Objects 6 (1974)).

[xvii] Portions of this Introduction, including the explanation of these two hypothetical entities, first appeared in Boyle, supra note 6.

With Christmas fast approaching, on December 8, the European Parliament wrapped up one of its biggest presents of the mandate: the AI Act. A landmark piece of legislation with the goal of regulating Artificial Intelligence while encouraging development and innovation. In sticking with the holiday theme, the last weeks of the negotiations have included everything from near-breakdowns of the discussions, not too dissimilar to the explosive dynamics of festive family gatherings, and 20+ hour trilogue meetings, akin to last-minute christmas shopping. But alas, it is done.

One of the key priorities for COMMUNIA was the issue of transparency of training data. In April, we issued a policy paper calling the EU to enact a reasonable and proportional transparency requirement for developers of generative AI models. We have followed the work up with several blogposts and a podcast, outlining ways to make the requirement work in practice, without placing a disproportionate burden on ML developers.

From our perspective, the introduction of some form of transparency requirement was essential to uphold the legal framework that the EU has for ML training, while ensuring that creators can make an informed choice about whether to reserve their rights or not. Going by leaked versions of the final agreement, it appears that the co-legislators have come to similar conclusions. The deal introduces two specific obligations on providers of general-purpose AI models, which serve that objective: an obligation to implement a copyright compliance policy and an obligation to release a summary of the AI training content.

In a leaked version, the obligation to adopt and enforce a copyright compliance policy reads as follows:

[Providers of general-purpose AI models shall] put in place a policy to respect Union copyright law in particular to identify and respect, including through state of the art technologies where applicable, the reservations of rights expressed pursuant to Article 4(3) of Directive (EU) 2019/790

Back in November, we suggested that instead of focussing on getting a summary of the copyrighted content used to train the AI model, the EU lawmaker should focus on the copyright compliance policies followed during the scraping and training stages, mandating developers of generative AI systems to release a list of the rights reservation protocols complied with during the data gathering process. We were therefore pleased to see the introduction of such an obligation, with a specific focus on the opt-outs from the general purpose text and data mining exception.

Interestingly, the leaked version contains a recital on which the co-legislators declare their intent to apply this obligation to “any provider placing a general-purpose AI model on the EU market (…) regardless of the jurisdiction in which the copyright-relevant acts underpinning the training of these foundation models take place”. While one can understand why the EU lawmakers would want to ensure that all AI models released in the EU market respect these EU product requirements, the fact that these are also copyright compliance obligations, which apply previously to the release of the model in the EU market, would raise some legal concerns. It is not clear how the EU lawmakers intend to apply EU copyright law when the scrapping and training takes place outside the EU borders without an appropriate international legal instrument.

The text goes on to require that developers of general-purpose AI models make publicly available a sufficiently detailed summary about the AI training content:

[Providers of general-purpose AI models shall] draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office

While we have previously criticized the formulation “sufficiently detailed summary” due to the legal uncertainty it could cause, having an independent and accountable entity draw-up a template for the summary (as we defended in here) could alleviate some of the vagueness and potential confusion.

We were also pleased to see that the co-legislators listened to our calls to extend this obligation to all training data. As we have said before, on the one hand introducing a specific requirement only for copyrighted data would add unnecessary legal complexity, since ML developers would first need to know which of their training materials are copyrightable, and on the other hand knowing more about the data that is feeding models that can generate content is essential for a variety of purposes, not all related to copyright.

We should also highlight that the co-legislators appear to have a similar understanding to ours in terms of how compliance with the transparency requirement could be achieved when the AI developers use publicly available datasets. In the leaked version there is a clarifying recital stating that “(t)his summary should be comprehensive in its scope instead of technically detailed, for example by listing the main data collections or sets that went into training the model, such as large private or public databases or data archives, and by providing a narrative explanation about other data sources used.”. When the training dataset is not publicly accessible, we maintain that there should be a way to ensure conditional access to the dataset, namely through a data trust, to confirm legal compliance.

Taking these amendments into account, the compromise found by the co-legislators manages to strike a good balance between what is technically feasible and what is legally necessary.

Merry Christmas!

The post An AI Christmas Miracle appeared first on COMMUNIA Association.