News Media Europe: Statement of concern regarding draft AI code of practice

European news publishers are deeply concerned about the third draft code of practice published last week by the European Commission. The draft includes a chapter about how AI providers must comply with EU copyright law which has unfortunately gone from bad to worse: it contains serious errors relating to the interpretation of AI and copyright law.

‍

As a participant, we are very concerned about the quality, independence and fairness of the consultation process. The draft the code raises questions about whether it is used to further a broader political agenda, motivated by the belief that undercutting intellectual property rules will boost competitiveness.

‍

The latest draft confirms our fears that no code is better than this code. The legal interpretations drafted by the chairs of the copyright working group are invariably restrictive and arbitrary, they deliberately overlook the application of general principles of copyright law. Repeated feedback pointing to this problem was shared but consistently ignored.

‍

The outcome is a draft code that is generally unclear and presents a diluted version of the EU’s AI Act that contradicts copyright law. It completely fails to meet its objectives insofar as copyright is concerned. As a result, the code risks creating backdoor legislation while rightsholders are still unable to verify if and how their works are used by AI systems, preventing them from being able to exercise their rights.

‍

The third draft code includes but is not limited to the following problems:

• Absence of general principle of lawful access and authorisation which underpins copyright law
• Inadequate language about the appropriateness of obligations of means whereas instead obligations of results should apply as evidenced by a large body of case law
• Absence of public transparency regarding copyright compliance policy of AI providers, which will make compliance assessment more difficult
• Absence of meaningful due diligence in the use of third party datasets by AI providers who are not required to engage in any fact finding exercise
• Arbitrary privilege of robots.txt protocol as means of exercising a reservation of rights despite collective call for flexibility by rightsholder representatives
• Arbitrary exclusion of non-protocol based means of exercising a reservation of rights, such as legal terms and conditions, widely used by rightsholders
• Absence of meaningful due diligence to prevent copyright infringing outputs, with no justification

‍

Consequently, the code’s copyright chapter is wholly inadequate and should not be endorsed as such by the European Commission and Member States. The enforcement of copyright law in the age of AI is a life-or-death question for European press publishers. The future of journalism is at stake.

‍

The relevant Commissioners must take corrective measures concerning the work of the AI Office, if it is not able address the serious problems linked to the code. The draft code, as it stands, is of no meaningful help to rightsholders. We stand ready to continue discussions to reach a workable compromise.

‍

Contact: Iacob Gammeltoft, Senior EU Affairs Manager

Original post
News Media Europe Statement of concern regarding draft AI code of practice