AI Model Developers Now Required to Disclose Information under EU Regulations
The European Union has taken a significant step in regulating Artificial Intelligence (AI) with the adoption of the EU AI Act in May 2024. Effective since February 2025, this comprehensive legal framework aims to ensure AI systems operating within the EU are trustworthy, safe, human-centric, and respect fundamental rights.
Transparency and Disclosure
The Act mandates transparency for many AI systems to help users understand when they are interacting with AI and its implications. Developers must disclose their AI systems' workings and the data they were trained on. The EU Commission has issued guidelines specifically clarifying these obligations for General-Purpose AI (GPAI) models, including notification duties for providers.
Intellectual Property Protection
While the EU AI Act does not provide detailed specifics on intellectual property (IP) provisions, it focuses on harmonized rules ensuring safety, rights protection, and trustworthy use, which may indirectly influence IP concerns related to AI-generated content and models. However, explicit mechanisms on IP protection under the AI Act are not detailed in the provided information.
Enforcement and Compliance
The EU AI Act is enforced through a phased approach, with several regulatory provisions becoming applicable at intervals up to 36 months. The Act's enforcement includes monitoring AI literacy obligations, risk-based system classification adherence, and prohibitions on unacceptable risk AI usage. The establishment of an EU AI Office serves as a central coordination point for enforcement and oversight.
Penalties for Violations
Details on penalties are not fully covered; however, non-compliance—especially related to high-risk or prohibited AI systems—likely entails significant legal and financial penalties. These penalties are designed to ensure effective enforcement, similar to other EU regulatory frameworks like GDPR.
New Rules for AI Model Providers
The EU has implemented new rules for AI model providers, effective from August 2, 2023. Developers of AI systems must now report the sources they used for their training data and disclose whether they automatically scraped websites for their training data.
Exemptions and Timelines
AI models that were on the market before August 2, 2025, will be controlled from August 2027. The EU Artificial Intelligence Authority will enforce the new AI rules from August 2026 for new models.
Controversies and Concerns
Several national and international associations of authors, artists, and publishers complain that the legislation does not sufficiently protect intellectual property. The Initiative for Copyright believes the measures in the legislation are ineffective in terms of intellectual property protection. Google has expressed concern about the new AI regulations.
Strengthening Copyright Protection
The new rules aim to strengthen copyright protection among other things. Developers must specify what measures they have taken to protect copyright. Operators of these AI systems are now required to disclose a contact point within the companies for rights holders.
Private Individuals and Legal Action
Private individuals can sue AI providers based on the AI Act. Particularly powerful AI models that could potentially pose a risk to the public must also document their security measures. Violations of the AI Act can result in fines of up to 15 million euros or three percent of the total global annual turnover.
In summary, the EU AI Act introduces a pioneering, risk-based regulation for AI with a strong focus on transparency, safeguarding fundamental rights, and ensuring human oversight, enforced through multi-level governance, and backed by mechanisms expected to include significant penalties for violations.
Technology companies developing artificial-intelligence (AI) systems within the European Union must disclose the workings and the data their AI systems were trained on, as mandated by the EU AI Act. Additionally, these companies must report the sources they used for their training data if they are AI model providers, effective from August 2, 2023.
