The AI Office has initiated a multistakeholder consultation to address the application of the AI Act's definition of an AI system and its list of prohibited practices. This effort aims to guide the European Commission in preparing guidelines before the Act's provisions come into effect on February 2, 2025. Unlike previous consultations, this one does not provide a draft for feedback but instead seeks clarification on existing definitions and identification of systems that might be prohibited. The consultation is open until December 11.

Additionally, the AI Office released a draft General-Purpose AI Code of Practice as part of the compliance process established by the AI Act. This draft outlines measures for transparency, risk identification and assessment, risk mitigation, and internal governance for general-purpose AI model providers. Key points include documentation requirements for transparency, a taxonomy of systemic risks including future-looking threats, and safety mitigations detailed in a Safety and Security Framework.

Participants in the Code of Practice process must submit their feedback by November 28.

In other developments related to the AI Act, there was a consultation on rules for forming a scientific panel of independent experts. Civil society responses emphasized enhancing independence requirements for selected experts. Moreover, appointing fundamental rights authorities across member states has been inconsistent despite deadlines; only seven countries have announced their nominees.

The European Data Protection Board (EDPB) held an event on November 5 to gather input for an upcoming opinion on personal data processing during AI model training and operation stages. Meanwhile, the AI Office will conduct webinars on November 28 and December 17 about the architecture of the AI Act and host a workshop on GPAI models' systemic risks on December 13.