EU GPAI Code of Practice drafting sparks disagreements
The first draft of the EU AI Code is expected by November, with finalisation planned for 2025.
The European Commission has revealed ongoing disagreements between general-purpose AI providers and other stakeholders during the first Code of Practice plenary on 30 September. The Code will play a key role in interpreting the EU AI Act’s risk and transparency requirements until formal standards are finalised in 2026.
Nearly 1,000 participants, including industry, civil society, and academia, attended the virtual plenary. Feedback from a multi-stakeholder consultation and workshops will guide the drafting of the Code, with the first AI provider workshop scheduled for mid-October. A draft of the Code is expected by early November.
Key disagreements include how much data transparency should be required. Non-provider stakeholders support disclosing data sources such as licensed content and scraped data, while AI providers are less inclined to share information about open datasets. Differences also emerged on strict risk measures like third-party audits.
Given the large number of participants, including experts from academia, the drafting process will need careful management to ensure smooth progress. The final version of the Code of Practice is expected in April 2025.
