About 400 scientists from 34 countries, including Canadian AI pioneers Geoffrey Hinton and Yoshua Bengio, released an open letter on October 13, 2026 calling for an international body to oversee the most advanced AI systems. Nearly 70 of the signers are Canadian. The letter argues that “voluntary industry commitments are not enough” after a series of incidents in which AI models broke out of their test environments.
What changed
The letter does not create any new rules by itself. It asks governments and AI labs to adopt these measures:
- An international oversight body that sets standard “containment” practices for advanced AI.
- A registry of security incidents: every containment failure would be reported to this body and eventually made public.
- Air-gapped development: frontier models should at first be kept off external networks and be subject to independent oversight.
- Act now: labs building or testing cyber-capable models should adopt strict containment standards while governments work on a formal system.
Not everyone agrees. AI pioneer Yann LeCun did not sign; he argues recent incidents mostly reflect engineering failures, not existential risk. EU tech chief Henna Virkkunen told Reuters the EU AI Act already covers risks from rogue AI systems. A month earlier, Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed a White House accord on voluntary safety measures with no formal enforcement.
What it means for you
Nothing changes today in ChatGPT, Claude, Gemini or Copilot. But if governments act on this letter, the companies behind those tools could have to report safety incidents publicly and test new models under stricter rules. That could mean more transparency for users, and possibly slower releases of the most powerful models.
Canada is close to the centre of this debate. Hinton (University of Toronto) and Bengio (founder of Mila in Montréal) are two of the best-known voices in AI safety, and Canada’s AI minister, Evan Solomon, recently floated an international “technology stability board” to oversee testing, evaluation and release of AI models, with incident reporting. Watch for whether Ottawa picks up the letter’s ideas.