GettyImages © FangXiaNuo
General-purpose AI model providers are the main addresses of the code, following the Article 56 of the AI Act, and thus have the opportunity to discuss the content of the code in workshops, in addition to working group discussions with all plenary participants.
The code of practice aims to shape the future development and governance of trustworthy AI in Europe. These workshops are crucial as they allow providers to discuss directly with the Chairs and Vice-Chairs during the four drafting rounds.
After an opening session with the four working groups that form the Code of Practice Plenary, this first workshop will cover:
- Systemic risk assessment, technical mitigation, and governance (working Groups 2, 3, and 4)
- Transparency and copyright-related rules (working Group 1)
The (Vice-)Chairs will lead the workshop discussions, asking guiding questions, for instance based on the providers’ recent consultation submissions. Discussions will be held under Chatham House rules. The AI Office will ensure transparency in these discussions by drawing up high-level meeting minutes which will be made available to all plenary participants.
The first draft of the Code is expected to be shared with all the participants of the plenary, including providers, in the week of 4 November.
Background
On 30 September, the European AI Office kicked off the process for drawing-up of the first code of practice for general-purpose AI models, under the AI Act. Nearly 1 000 attendees including general-purpose AI model providers, downstream providers, industry, civil society, academia, and independent experts, took part in the online plenary to help develop the code of practice.
Providers of general-purpose AI models will be invited to special workshops with the Chairs and, when needed, the Vice-Chairs of the code of practice. These workshops will allow them to contribute to each stage of the drafting process, in addition to their participation in the plenary.
The code of practice aims to support the correct application of the AI Act's provisions for general-purpose AI models, including those related to transparency, copyright, systemic risk taxonomy, risk assessment, and mitigation measures.
The drafting process involves four working groups, each meeting three times to discuss various drafts. It is overseen by Chairs and Vice-Chairs, independent experts selected following a call for expressions of interest.
The final version of the code of practice will be published and presented at a closing plenary, foreseen to April 2025. Further information on the development of the first General-Purpose AI code of practice is available online.