Yoshua Bengio calls for mandatory audits of AI developers by evaluators they do not choose
X · Personal post
Read the source (opens in a new tab)- What happened
- Yoshua Bengio posted on X on November 10, 2026, in French and English, welcoming a meeting President Emmanuel Macron convened that day with heads of government, frontier AI company leaders and researchers.
- Key points
- "Model developers should undergo mandatory, independent audits by evaluators they do not choose," Bengio wrote.
- He says developers should be "held accountable for demonstrating that their models are safe" before the models can pose a grave danger, which he says can exist before deployment.
- He calls for treating AI like other technologies that "can deeply transform and threaten our societies," with "intentional and rigorous governance frameworks."
- He says the meeting discussed "the growing risks of advanced AI, including dangerous loss of control and misuse of AIs."
- Evidence
- A researcher's own post stating a policy position; it cites no new evidence. Bengio's text does not name the meeting's outcome or say how such audits would be set up.