AI Model Providers to Adopt Transparency under EU Regulations
The European Union has introduced new rules for AI models, effective from August 2, 2023. These rules apply to companies like ChatGPT and Gemini, aiming to strengthen copyright laws and ensure transparency in AI systems. However, AI associations and 38 global creators' organizations have expressed strong concerns that the EU AI Act fails to adequately protect intellectual property rights.
Under the new regulations, developers must now report which sources they used for their training data and whether they automatically scraped internet sites for the same. They must also detail the measures taken to protect intellectual property. Operators of AI systems must disclose how their systems function and the data used to train them. Highly capable AI models that could potentially pose a risk to the public must document their safety measures.
The European AI Agency will begin enforcing these new rules for new models from August 2026, and for models released before August 2, 2025, from August 2027. Private individuals can sue providers based on the AI Act, and rights holders will have a point of contact within the companies. The EU Commission can impose fines on companies that violate the AI Act, with penalties reaching up to 15 million euros or three percent of the company's global annual turnover.
Despite these efforts, AI associations and creators' organizations have criticized the legislation for not adequately protecting intellectual property rights. In a joint statement, these groups—including authors, performers, publishers, and producers—argue that the AI Act’s current framework insufficiently balances the interests of rights holders against the development and deployment of AI systems. They claim that the legislation represents a failure to deliver on the EU's own commitments regarding IP protection within AI regulation.
Their dissatisfaction particularly pertains to elements such as the Code of Practice, the GPAI Guidelines, and the requirement under Article 53 for disclosure of a sufficiently detailed summary of training data. These associations argue that the AI Act’s current framework undermines protections for creators' intellectual property in AI training and usage contexts.
Google, like many others, has also shown concerns about the new legislation. As the AI landscape continues to evolve, it remains to be seen how these new rules will impact the industry and creators alike. The EU AI Act, adopted in May 2024, marks a significant step towards regulating AI, but the debate over intellectual property protection will undoubtedly continue.
[1] Source: Joint statement by AI associations and 38 global creators' organizations, dated July 31, 2023.
- The new EU AI regulations, effective from August 2, 2023, require developers to detail the measures taken to protect intellectual property rights, addressing the concerns raised by AI associations and creators' organizations.
- In a joint statement, AI associations and 38 global creators' organizations criticized the EU AI Act for not adequately protecting intellectual property rights, particularly focusing on the Code of Practice, the GPAI Guidelines, and the requirement under Article 53 for disclosure of a sufficiently detailed summary of training data.
