On 18 July 2025, the European Commission published guidelines on the scope of the obligations for providers of general-purpose AI (GPAI) models under the AI Act. These guidelines aim to assist actors across the AI ecosystem in determining whether the obligations introduced by the AI Act apply to them, and if so, what is expected in terms of compliance. They are intended to provide legal certainty and support innovation.
These obligations will apply from 2 August 2025, pursuant to the AI Act. The guidelines do not constitute a binding legal instrument but reflect the Commission’s official interpretation and are expected to guide regulatory enforcement.
Key Elements of the Guidelines
- Clarification of Definitions: The guidelines set out clear technical criteria to determine whether an AI model qualifies as a “general-purpose” model. This helps developers and downstream users understand whether the AI Act’s specific obligations for GPAI models are applicable.
- Pragmatic Compliance Approach: The guidelines adopt a practical stance. Notably, they specify that only entities making significant modifications to GPAI models are subject to the obligations applicable to providers. Entities making minor or limited changes are not considered providers for the purposes of these obligations.
- Exemptions for Open-Source Models: The Commission clarifies the conditions under which providers of open-source GPAI models may be exempted from certain obligations. This approach seeks to encourage openness, transparency, and innovation while preserving safeguards.
Purpose and Legal Significance
The guidelines offer an interpretation of how the Commission understands the application of key terms and provisions in the AI Act. In doing so, they complement the General-Purpose AI Code of Practice, a non-binding and practical tool which provides operational guidance on how providers may fulfil their obligations under the Regulation.
The guidelines aim to enable developers, deployers, and other stakeholders in the AI value chain to operate with clarity, legal certainty, and confidence.
Timeline for Application and Enforcement
- From 2 August 2025: The obligations for providers of GPAI models will apply. GPAI models placed on the EU market from this date forward must comply with the applicable requirements. Providers of systemic-risk GPAI models are under a legal obligation to notify the AI Office. All GPAI providers are encouraged to cooperate informally with the AI Office’s technical team, with additional support available to those who are signatories of the Code of Practice.
- From 2 August 2026: The Commission’s enforcement powers under the AI Act become effective. It will begin verifying compliance and may impose sanctions, including fines, in case of violations.
- By 2 August 2027: Providers of GPAI models placed on the market before 2 August 2025 must ensure full compliance with the AI Act by this date.
Recommendations for Providers
The guidelines call upon providers of GPAI models to:
- Carefully assess whether their models fall within the definition of GPAI models under the AI Act;
- Determine whether their models present systemic risk requiring notification to the AI Office;
- Review and adapt internal governance, documentation, and compliance systems;
- Engage proactively with the AI Office to ensure timely alignment with the applicable legal requirements.
Providers that are required to notify the Commission of a GPAI model with systemic risk should do so by contacting the AI Office at:
EU-AIOFFICE-GPAI-SR-PROVIDERS@ec.europa.eu
The AI Office has also made available practical guidance for notification.
The full guidelines, published on 18 July 2025, are available at the following link:
European Commission – Guidelines for Providers of General-Purpose AI Models (18 July 2025)
