AI ACT: Coalition of authors, performers and other rightsholders issue statement on the GPAI Code of Practice

A key objective of the AI Act is to give authors, performers and other rightsholders tools to exercise and enforce their rights by requiring General-Purpose AI (GPAI) providers to put in place measures to comply with EU copyright law and provide a sufficiently detailed summary of the content ingested and used for training. This was recognised by the EU legislator as a key means to support the development of AI in Europe and protect the EU’s creative sectors that drive its cultural and economic landscape.

However, the third draft of the GPAI Code of Practice represents yet another step away from achieving this objective. It creates legal uncertainty, misinterprets EU copyright law and undercuts the obligations set out in the AI Act itself. Rather than providing a robust framework for compliance, it sets the bar so low as to provide no meaningful assistance for authors, performers and other rightsholders to exercise or enforce their rights. Even more fundamentally, it would also not ensure that GPAI providers comply with either EU copyright law or the AI Act.

We have participated constructively in the drafting process and provided substantive comments to the previous drafts of the Code of Practice, underlining severe deficiencies affecting all creative sectors. However, the drafters of the Code have largely ignored or dismissed those comments. Regrettably, the third draft does not meet the adequacy requirement provided by the EU AI Act, and, therefore, should not be approved without substantial improvements. No Code would be better than the fundamentally flawed third draft.

The third draft further reduces the obligation to ensure compliance with EU copyright rules and the AI Act itself. In several places it states that GPAI providers merely need to make “reasonable efforts” to ensure compliance with EU copyright law. However, “reasonable efforts” or mere encouragement for GPAI providers to take measures is not sufficient to ensure compliance with EU copyright law or the AI Act.

The draft waters down GPAI providers’ responsibility to undertake proper due diligence to ensure that the third-party datasets they use to train their models do not infringe copyright. Not only would this eliminate any meaningful due diligence obligations to comply with EU copyright law and the AI Act, but it would also risk guiding GPAI providers towards copyright infringements.

The third draft continues to render meaningless the right of authors, performers and other right holders to choose how they reserve their rights. It also fails to provide meaningful guidance on what GPAI providers must do to comply with such reservations. The draft further suggests that GPAI providers could comply with the “lawful access” requirement solely by adhering to technical access control measures, combined with “reasonable efforts” to exclude only a narrow and incomplete set of piracy sites from training. Given that lawful access is a mandatory requirement for eligibility to the TDM exceptions, if they apply, this measure fails to meet the standard set by substantive EU copyright law. Moreover, it not only misinterprets EU copyright law, but also does not reflect the reality of how pirated content is accessed and distributed online.

The third draft removes entirely the requirements to demonstrate transparency regarding GPAI providers’ compliance with the right reservation mechanisms; GPAI providers are not required to disclose whether or how they comply with the rights reservations expressed by authors, performers and other rightsholders. Despite strong opposition and comprehensive explanations on its inadequacy, robots.txt is still considered as the only method GPAI providers must recognise and respond to, while other reservation methods are ignored or treated as optional, in direct contradiction with EU law.

Furthermore, the measure for GPAI providers to introduce a copyright complaints process only requires the introduction of a mechanism to lodge complaints in relation to the Code of Practice alone and without any reference to the measures GPAI providers should take to resolve them, rendering it an empty gesture rather than a meaningful enforcement tool.

The Code of Practice should provide proper measures to facilitate and persuade GPAI models to respect the two basic principles of copyright law: they should seek prior authorisation and abstain from unauthorised uses of copyrighted material. The Code of Practice should also make clear, as explained also by EVP Virkkunen in her recent response to a parliamentary question, that the obligations in the AI Act apply where a GPAI provider places their model into the EU market, regardless of where they are established or where the training of that model took place.

The Code of Practice must also be coupled with an effective template for the “sufficiently detailed summary of the content used for training”, enabling authors, performers and other rightsholders to effectively exercise or enforce their rights. Meaningful, actionable transparency on the content used for training and other purposes, and its disclosure, is not only feasible but can be achieved with little technical or financial effort. Misusing trade secret law to hide infringements would render the obligation meaningless and obstruct the fundamental right of rightsholders to exercise and enforce their rights.

The intention of the EU’s AI Act was to enable the development of responsible AI and ensure that the cultural and creative sectors can continue to grow across European territories. Entirely counter to this, the third draft of the GPAI Code of Practice sets an implementation standard which undermines the objectives of the AI Act, contravenes EU law and ignores the intention of the EU legislator.

The third draft is therefore completely unacceptable. We cannot support the GPAI Code of Practice in its current version. No Code would be better than the fundamentally flawed third draft.
—