Commission Issues Guidelines on General-Purpose AI Model Obligations

The European Commission has published guidelines clarifying the obligations for providers of general-purpose AI (GPAI) models under the AI Act, with these requirements set to apply from 2 August 2025. The guidelines detail technical criteria for when an AI model qualifies as “general-purpose,” providing much-needed clarity for developers and legal teams. Importantly, only providers making significant modifications to GPAI models must comply with these obligations, while those making minor changes are excluded.

Additionally, the guidelines address exemptions for open-source AI models, outlining under what circumstances these providers are exempt from certain requirements. This is intended to encourage transparency and innovation in the sector without imposing unnecessary regulatory burdens on open-source initiatives.

The guidelines serve as an interpretative tool, helping stakeholders understand key terms and obligations under the AI Act. They complement the General-Purpose AI Code of Practice, a voluntary framework that offers practical steps for compliance. Although not legally binding, these guidelines will inform enforcement actions and are the result of extensive public consultation with stakeholders across the AI ecosystem.

Enforcement of these obligations will begin on 2 August 2026, with fines for non-compliance. Providers of GPAI models launched before 2 August 2025 must achieve compliance by 2 August 2027. Providers of models posing systemic risk are required to notify the AI Office and can access practical guidance through official EU channels. Legal practitioners and AI stakeholders are advised to review these guidelines, assess their compliance obligations, and engage proactively with the AI Office.