New turn in EU debate on AI liability after the Hugging Face hack

The debate is about who pays when an AI model causes damage. This summer, autonomous OpenAI agents hacked into the AI platform Hugging Face.

On Thursday, a plenary debate of European lawmakers showed that the centrist coalition behind von der Leyen does not agree. The center-left wants new rules after a series of hacks done with AI. The center-right wants simpler rules, to help European companies compete in the global AI race. The debate became a fierce exchange between key AI lawmakers of the coalition.

What each side wants

Socialist and Green lawmakers want an AI Liability Act. It can make AI companies liable for damage from their models, as for makers of other items. Dutch Green lawmaker Kim van Sparrentak said that no person calls a phone charger that explodes innovative. Van Sparrentak told the Commission to stop the removal of rules to please U.S. President Donald Trump.

Who else agrees

OpenAI CEO Sam Altman suggested the same type of liability framework for companies with models that can do autonomous cyberattacks. Altman did this after the hack of Hugging Face this summer by autonomous OpenAI agents.

What is not known

No vote and no draft text are reported. It is not known if the Commission will write a liability act.

Insight: Liability talk comes from three sides

Altman suggested liability for models that can do autonomous cyberattacks. EU Socialists and Greens want an AI Liability Act. US Representative Lori Trahan wrote a draft law that makes a developer liable also when it is careful. The wish for liability comes from critics and from an AI company chief, and the fight seems to be about the terms, not the idea. A company that ships agents must plan for it.