⚡ Quick Summary
Published by the European Commission, this Communication to the Commission of 18 July 2025 (C(2025) 5045 final) asks the College to approve the content of a draft Communication containing Guidelines on the scope of the obligations for general-purpose AI models under Regulation (EU) 2024/1689 (the AI Act). It recalls that the AI Act entered into force on 1 August 2024 and that Chapter V lays down obligations for providers of general-purpose AI systems in two categories: models with and without systemic risk.
The draft Guidelines, mandated by Article 96(1), are meant to increase legal clarity ahead of the rules' entry into application from 2 August 2025. They clarify key concepts – what is a 'general-purpose AI model', a 'provider of a general-purpose AI model', the 'placing on the market of a general-purpose AI model' – and how to estimate the computational resources used for training such a model. They also set out how the AI Office will support and work with providers, an approach the document calls pragmatic and practical.
The Communication places the Guidelines in a package that also contains the General-Purpose AI Code of Practice, an adequacy assessment by the Commission and the AI Board, a template for transparency on training data and a template for notifications from providers of models with systemic risk. Formal adoption follows once all language versions are available; the draft Guidelines are enclosed as an Annex.
🧩 What’s Covered
All content comes from the two-page Communication; the draft Guidelines referred to as an Annex were not part of the text provided.
- Context and entry into force: states that Regulation (EU) 2024/1689 entered into force on 1 August 2024 and describes its twin aim of promoting innovation in and uptake of AI in the Union while ensuring a high level of protection of health, safety and fundamental rights, "including democracy and the rule of law".
- Chapter V scope: obligations for the providers of general-purpose AI systems, "divided into two categories: general-purpose AI model with and without systemic risk".
- Legal mandate: Article 96(1) requires the Commission to develop guidelines on the practical implementation of the AI Act, the basis for the annexed draft Guidelines.
- Purpose of the draft Guidelines: to increase legal clarity and give insights into the Commission's interpretation of the provisions on general-purpose AI systems, in light of their imminent entry into application from 2 August 2025.
- Concepts to be clarified: 'general-purpose AI model', 'provider of a general-purpose AI model', 'placing on the market of a general-purpose AI model', and how to estimate the computational resources used for training such a model.
- AI Office support: the draft Guidelines explain how the AI Office will support and work with providers who must comply with the general-purpose AI rules.
- Accompanying package: the General-Purpose AI Code of Practice and the adequacy assessment by the Commission and the AI Board, the Template for transparency on training data, and a template for notifications that providers of models with systemic risk submit to the AI Office.
- Approval and adoption procedure: approval of the content of the draft Communication now, formal adoption by the Commission at a later date once all language versions are available, with applicability only from that moment.
💡 Why it matters?
For providers of general-purpose AI models and the teams supporting them, this Communication indicates what the Commission intends to clarify before the Chapter V obligations become directly applicable on 2 August 2025. It identifies the questions the forthcoming Guidelines will answer – who counts as a provider, what placing on the market means, how training compute is estimated – and names the companion instruments (Code of Practice, training-data transparency template, systemic-risk notification template) that providers will work with.
It also states plainly that the annexed Guidelines are not yet applicable, so compliance planning has to account for the adoption and translation timeline rather than treating the draft text as binding.
❓ What’s Missing
The Communication is a short procedural text. The draft Guidelines it approves are annexed but were not part of the text provided, so no substantive interpretation, threshold or duty appears here. It gives no figure for the computational resources threshold, no fixed timetable beyond "a later date" for formal adoption, and no detail on the content of the Code of Practice, the training-data transparency template or the systemic-risk notification template. Questions of enforcement, penalties, or how the Guidelines relate to other legal instruments are not addressed, and the whole document is framed for the Union legal order.
👥 Best For
Best for compliance and public-affairs teams at providers and deployers of general-purpose AI models who need to track what the Commission will clarify before 2 August 2025, and for policy analysts mapping the AI Act's guidance package, its components and its adoption timeline.
📄 Source Details
Approval of the content of the draft Communication from the Commission – Guidelines on the scope of the obligations for general-purpose AI models established by Regulation (EU) 2024/1689 (AI Act), Communication to the Commission C(2025) 5045 final, European Commission, Brussels, 18 July 2025. Two pages, in English; no named authors are given. No URL is printed in the document. The extracted text covers the full two-page Communication; the draft Guidelines referred to as an Annex were not included in the extraction.
