Commission calls on tech companies, experts and activists to help write rules for powerful AI

The AI Office, an arm of the European Commission overseeing the implementation of the bloc’s artificial intelligence rulebook, is asking tech companies, academics and activists to help draft rules for powerful general-purpose AI models.
In a call for expression of interest published today, the AI Office said that AI developers, industry organizations, civil society, academics and other experts can apply to join in the process by Aug. 25.
Those selected will compose a “plenary” working with the AI Office to thrash out how AI labs should practically comply with the rules established by the AI Act on governance, AI “systemic risks,” and transparency on the use of copyrighted content to train AIs.
The plenary’s first and closing sessions are slated for September 2024 and April 2025, respectively; the AI Office expects the plenary to hold three virtual “drafting rounds.” The talks will likely be overseen by an external contractor.
The process will result in a code of practice AI companies could follow to easily “demonstrate compliance” with the regulation, according to the AI Office’s call.
The plenary’s code will need the green-light of the AI Office itself and of EU government experts — which have three months to decide whether to approve or discard it and have the Commission draft rules instead. The AI Act’s rules for general-purpose AI enter into force in August 2025.
The plenary will be subdivided into four working groups, focused on: transparency and copyright rules; assessment of systemic risks; systemic risks mitigation; and governance.
AI companies will also attend workshops with the working groups’ chairs to set the discussions’ general direction. Other plenary members will have access to the workshops’ minutes.
The AI Office also opened a consultation today on trustworthy AI, which will feed into the code of practice’s first iteration.