The plenary vote to adopt the own-initiative (INI) report initiated by Axel Voss brings to a close a process that began in June 2025. While many aspects of the report improved during the committee stage and the text itself does not have immediate legal consequences, the overall outcome remains disappointing. The report is still incoherent and fails to offer a clear path forward. It also continues to include a number of problematic provisions, particularly concerning press publishers’ and media content, which we had hoped the European Parliament would address before proceeding to the final vote.
No further amendments were tabled after the committee vote, meaning the report adopted in plenary is identical to the version we analysed in our previous blog post.
Some improvements
Despite our overall concerns, the final report does contain several improvements compared to the original draft. Most importantly, the report now recognises that the text and data mining (TDM) exceptions apply to AI training, thereby reducing legal uncertainty and aligning the text with the existing EU legal framework. This clarification is significant, given ongoing debates about the legality of using copyrighted works to train AI systems.
Another area of improvement concerns transparency obligations for commercial AI developers. Instead of calling for an irrebuttable presumption of use, the report now refers to a rebuttable presumption, which is a more proportionate and legally sound approach.
Finally, the adopted text includes some modest improvements with regard to users’ rights. It calls on the Commission to explore measures addressing copyright infringements arising from generative AI outputs, while ensuring that such measures do not prevent the generation of non-infringing content.
Problematic language on press publishers’ and media content
At the same time, the report contains provisions that are deeply problematic. In particular, its language on media content can be interpreted extremely broadly. The text calls on the European Commission to explore extending both the press publishers’ right and broadcasting rights in order to grant press and news media rightsholders “full control” over the use of their content for AI training and related purposes. According to the report, such uses would require “explicit consent.”
This framing risks undermining the existing balance in EU copyright law. In particular, it could be interpreted as excluding press and news media content from the scope of the TDM exceptions—an outcome that would significantly restrict the ability to train AI systems on publicly available information.
What is the impact of the report
It is important to recall that own-initiative reports are non-legislative procedures. They do not require the European Commission to propose legislation or take specific action.
Given the report’s mixed and at times internally contradictory positions, it is difficult to see how the Commission could translate it into a coherent legislative initiative. Nevertheless, the report does carry some political significance. Most notably, it underscores that the European Parliament supports the interpretation, also reflected in the AI Act, that the TDM exceptions apply to AI training. This confirmation is particularly relevant for researchers as well as to developers working on public-interest AI systems, who depend on legal clarity regarding the use of AI training data.
