Second Draft of the General-Purpose AI Code of Practice Released for Feedback
This article was collaboratively crafted by humans and AI, blending insights and precision to create a piece for your benefit. Enjoy!
Independent experts have unveiled the second draft of the General-Purpose AI Code of Practice, incorporating feedback from the first draft published on 14 November 2024.
This revised draft reflects the insights and recommendations of a broad range of stakeholders, including EU Member State representatives, European and international observers, and over 1,000 participants from the Code of Practice Working Groups and Provider Workshops.
The second draft is shaped by discussions from Working Group meetings held in the week of 18 November 2024, where participants provided verbal feedback, engaged in interactive polls, and had opportunities to ask questions and offer perspectives. A summary of these discussions was presented to the Plenary. In addition, 354 written submissions on the first draft were received via a dedicated feedback platform. These, along with feedback from AI model providers gathered in specialized workshops, were carefully considered in the updated draft.
The Chairs and Vice-Chairs also took part in a series of interinstitutional meetings to present the first draft and share progress:
- On 27 November 2024, they presented the first draft to the European AI Board Steering Group on General-Purpose AI.
- On 10 December 2024, they briefed the AI Board, a body established under the AI Act comprising representatives from all 27 EU Member States.
- On 11 December 2024, they updated the joint IMCO-LIBE Working Group of the European Parliament on the implementation and enforcement of the AI Act.
Status of the Second Draft
The General-Purpose AI Code of Practice is a guiding document for AI model providers, outlining how they can demonstrate compliance with the AI Act throughout their models' lifecycle. The second draft builds upon previous work with the goal of creating a “future-proof” Code, especially relevant for models released after 2 August 2025, when new regulations for general-purpose AI models will take effect.
The first part of the Code addresses transparency and copyright obligations for all AI model providers, with some exemptions for open-source providers, in line with the AI Act. A proposal for an adequate summary of training data content will be presented by the AI Office in early 2025, incorporating feedback from the Code of Practice participants before the Commission officially adopts it.
The second part of the Code focuses on advanced AI models that could pose systemic risks, as defined in Article 51 of the AI Act. It outlines measures for risk assessment and mitigation, including model evaluations, incident reporting, and cybersecurity obligations. The structure of this section has been updated to focus on objectives, commitments, and measures, replacing the previous draft’s goals, measures, and sub-measures. Preliminary Key Performance Indicators (KPIs) have also been included.
The draft emphasizes the need for flexibility in adapting to the rapidly evolving AI technology landscape. It underscores the importance of developing comprehensive ecosystems for AI governance and risk management, while maintaining proportionality in applying these measures—especially in consideration of the size and capabilities of the AI model provider.
Next Steps
In addition to the written feedback received by 15 January 2025, verbal discussions on the second draft with the Chairs and Vice-Chairs are planned, in alignment with the AI Office’s tentative timeline.
The upcoming Working Group meetings will take place during the week of 13 January 2025:
- 13 January: WG3 on Technical Risk Mitigation for Systemic Risk
- 14 January: WG1 on Transparency
- 15 January: WG2 on Risk Assessment for Systemic Risk
- 16 January: WG4 on Governance Risk Mitigation for Systemic Risk
- 17 January: WG1 on Copyright-Related Rules
Workshops with AI model providers and Member State representatives are scheduled for the weeks of 20 and 27 January 2025.
The third draft of the Code of Practice is expected to be released during the week of 17 February 2025.
This Q&A, provided by the AI Office, outlines the regulatory approach to general-purpose AI under the AI Act.
Written by Dev Anand from Funnel Fix It Team
