AI Model Providers Face Request for Open Disclosure under New EU Regulations
The European Union has introduced new rules for General-Purpose AI (GPAI) providers, effective August 2, 2023, as part of the EU AI Act. These regulations aim to strengthen copyright protection and promote transparency in AI development.
Under the new regulations, GPAI model providers are required to report the data used to train their models while respecting intellectual property rights, confidential business information, and trade secrets. The provisions, particularly Article 53(1)(c) and (d), are designed to facilitate holders of copyright and related rights in exercising and enforcing their rights under EU law.
However, several European creative and cultural organizations, including CISAC, ICMP, IFPI, and IMPALA, have criticized the AI Act’s implementation measures, particularly the GPAI Code of Practice, Guidelines, and Template for disclosure under Article 53. They argue that the current measures are inadequate and fail to deliver meaningful protection of intellectual property rights in the context of generative AI.
The criticism highlights that the feedback from the primary beneficiaries—creators and rightsholders—has been largely ignored, resulting in protections that favor AI model providers who continue to infringe copyrights by using large-scale scraping of protected works without proper licensing or compensation. This situation is seen as a missed opportunity and potentially damaging to the competitive position of traditional cultural and creative sectors in Europe.
Developers must also specify what measures they have taken to protect copyright and disclose how their systems work and the data they were trained on. There is no obligation to name specific data sets, domains, or sources in the legislation. Notably, Google has expressed concerns regarding the new rules.
In addition, the new guidelines require AI operators to report whether they automatically scraped websites for their training data. Models that came onto the market before August 2, 2025, will be controlled from August 2027. Particularly powerful AI models that could potentially pose a risk to the public must also document safety measures.
The European Artificial Intelligence Authority will have control over the new models from August 2026. In case of violations, fines of up to 15 million euros or three percent of the company's total global annual turnover may be imposed. Private individuals can sue providers based on the AI Act.
The Initiative for Copyright finds the measures ineffective due to the lack of obligation to name specific data sets, domains, or sources. The legislation does not adequately protect intellectual property according to several national and international associations of authors, artists, and publishers.
Despite the controversy, the new rules mark a significant step towards ensuring accountability and transparency in AI development while addressing concerns about copyright infringement. The European Artificial Intelligence Authority will play a crucial role in enforcing these new regulations and ensuring they are implemented effectively.
- Despite the ongoing criticism from several creative and cultural organizations, the new AI regulations in the EU, effective August 2, 2023, require AI operators to specify measures for protecting copyright, disclose the data used to train their models while maintaining intellectual property rights, and report deletions or modifications, if any, to these data.
- In the context of the EU AI Act, the new rules for GPAI providers seek to strengthen copyright protection by penalizing violators with fines of up to 15 million euros or three percent of their total global annual turnover, while the European Artificial Intelligence Authority oversights the implementation and enforcement of these regulations.
