France, Germany and Italy have agreed on a proposal to regulate AI in the European Union. It pushes for mandatory self-regulation rather than binding rules for foundational models without an initial sanctions regime.
The three governments support voluntary rules for large and small AI providers in the EU that would be binding if companies sign up to them, according to a joint paper seen by Reuters. There will be no initial sanctions.
The European Parliament's AI Act, unveiled in June, aims to manage the risks of AI applications and prevent discriminatory effects while promoting innovation.
Initially, the code of conduct was proposed to be binding only on major AI providers, mainly from the US. However, the three EU governments expressed concern that this could reduce trust in smaller European providers and lead to fewer customers.
Regulation of applications, not models
France, Germany, and Italy also argue that horizontal rules for foundation models would run counter to the technology-neutral and risk-based approach of the AI Act.
They propose that developers of foundation models define model cards that include information about the model's capabilities and limitations and that an AI governance body develop guidelines and review the use of model cards.
"We need to regulate the applications and not the technology if we want to play in the top AI league worldwide," Germany's Digital Minister Volker Wissing told Reuters.
In the future, a European authority could be set up to monitor compliance with the standards if violations of the code of conduct are detected after a certain period of time, Reuters reports. According to the non-paper, seen by Euractiv, "any suspected violation in the interest of transparency should be made public by the authority."
The European Commission, the European Parliament, and the EU Council are currently negotiating the bloc's stance on AI regulation. To learn more about the EU AI Act, see our deep dive here.