EU GPAI taskforce discusses systemic risks and copyright transparency

The fourth meeting of the Signatory Taskforce under the EU’s General-Purpose AI (GPAI) Code of Practice focused on the implementation of the Code’s Safety and Security and Copyright chapters.

During the meeting on 17 July, the European Commission’s AI Office discussed how information on the real-world use and behaviour of GPAI models could support post-market monitoring and systemic risk assessments. Such information could complement pre-deployment evaluations and help providers identify and manage risks after models have been released.

The taskforce also examined so-called marginal-risk clauses in safety and security frameworks. These clauses address situations in which a provider might consider changing its safety approach in response to competitors deploying increasingly capable but potentially unsafe models. The AI Office stressed that such provisions should only be used in exceptional circumstances and subject to appropriate safeguards.

The discussion on copyright focused on obligations requiring GPAI providers to provide greater transparency about the web crawlers they use and how those crawlers respect rights reservations expressed through mechanisms such as robots.txt. Signatories are also expected to make information about relevant compliance measures publicly available and provide a way for affected rightsholders to receive automatic notifications when that information changes.

The meeting concluded with an update on ongoing initiatives related to the GPAI Code of Practice and a discussion on the organisation of future taskforce meetings.