The European AI Office has published the first draft of the General-Purpose AI Code of Practice, developed by independent experts, which will guide the development of ‘trustworthy’ AI models.

The final text will detail how providers of general-purpose AI models may comply with their obligations under the AI Act.

The Code will be drafted in four rounds until April 2025, with the initial draft set to be discussed with around 1,000 stakeholders, member state representatives and international observers, in four dedicated working groups next week.

The first draft provides includes several open questions inviting feedback and further development by stakeholders.