The General Purpose AI Code of Practice is a forward-looking framework designed to guide the responsible development and deployment of artificial intelligence solutions. The first-of-its-kind document will detail the AI Act rules for providers of general-purpose AI models and general-purpose AI models with systemic risks. The European AI Office is facilitating the drawing-up of the Code, chaired by independent experts, involving nearly 1000 stakeholders, as well as EU Member States representatives, and European and international observers.
In December, a group of independent experts present the second draft of the General-Purpose AI Code of Practice. This document will serve as a “future-proof” Code, especially for models released after August 2, 2025, when the new rules for general-purpose AI models take effect.
The first section of this draft outlines the transparency and copyright requirements all general-purpose AI model providers must meet, with specific exemptions for certain open-source models by the AI Act. A proposal for a standardized summary of training data content is expected from the AI Office early next year, and the Office plans to integrate feedback from Code of Practice participants prior to the Commission’s formal adoption of the template.
The second section focuses on a select group of providers offering the most advanced general-purpose AI models that may pose systemic risks, as identified by the classification criteria in Article 51 of the AI Act. It describes measures for systemic risk assessment and mitigation, including model evaluations, incident reporting, and cybersecurity obligations. The structure of the Code has been updated to highlight objectives, commitments, and measures—replacing the previous arrangement of goals, measures, and sub-measures—and now includes initial examples of Key Performance Indicators (KPIs).
Another key goal of this draft is to clearly outline the obligations arising from the AI Act, further detailing how these can be fulfilled through concrete actions that the Signatories pledge to undertake.
Given the rapid pace of technological advancement, the Code stresses the importance of balancing clear commitments with the flexibility needed as technology evolves. It also underscores the necessity of developing ecosystems for AI governance and risk management. The Chairs note that the second draft remains a work in progress, focused primarily on clarifications, additional details, and proportionality (e.g., aligning obligations with the size of the AI model provider). Future iterations will aim to ensure that all sections of the Code integrate seamlessly, are easily understandable, and include further refinements.
The third draft of the Code of Practice is expected to be published during the week of February 17, 2025.
To learn more, download the Second Draft of the General Purpose AI Code of Practice here.
