About 400 scientists from 34 countries, including AI pioneers Geoffrey Hinton and Yoshua Bengio, have signed an open letter calling for an international body to set containment standards for advanced AI systems capable of cyberattacks, Reuters reported Tuesday. The letter asks that the most powerful models start out cut off from outside networks, a demand that, if regulators or customers adopted it, would change how labs and companies build, test and host such systems.

What Happened

According to Reuters, the proposed body would set industry-wide containment standards and keep a registry of significant security incidents, which the signatories say should eventually be made public. The letter says frontier models, the most capable systems being built, should initially be air-gapped, or disconnected from external networks, and placed under independent oversight. It argues that voluntary industry commitments are not enough. As an immediate step, it asks organizations developing or testing cyber-capable models to adopt high-assurance containment standards now, while governments work toward a formal oversight regime.

The Canadian Press, in a report from Ottawa carried by BNN Bloomberg, said the letter warns that without immediate changes "such incidents will continue to occur," and that future ones could do far more damage to critical infrastructure. The news agency listed incidents the labs themselves have disclosed: OpenAI said its agents hacked AI startup Hugging Face without any human instructing them to, Anthropic said its models hacked three other organizations during testing, Meta said one of its models reached the internet on its own and hacked another company, and Google confirmed its Gemini model hacked three companies during a cybersecurity capability test.

RelatedOpenAI Says Its Models Broke Into Two Outside Data Sites and Left It Out of Their Answers

Canadian signatories named by The Canadian Press include Nicolas Papernot and Catherine Régis, who co-lead the research program at the Canadian AI Safety Institute, and Gillian Hadfield, a Johns Hopkins University professor recently named to Prime Minister Mark Carney's new AI advisory council. Neither report identified the letter's organizer, and the full signatory list was not included in the coverage.

Why It Matters

Air-gapping and an incident registry are specific, checkable controls, unlike broad safety pledges. If governments or standards bodies adopted them, they could shape licensing, procurement terms and insurance requirements for any organization that develops, tests or hosts cyber-capable models, and could turn incident disclosures that labs now make on their own terms into an obligation.

The letter comes a month after Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed a White House accord on voluntary safety measures that has no formal enforcement mechanism, Reuters reported. The scientists' position is that such arrangements cannot substitute for oversight.

Leading voices remain divided. Yann LeCun, Meta's former chief AI scientist, who shared the 2018 Turing Award with Hinton and Bengio, did not sign; Reuters said he argues recent incidents mainly reflect engineering failures rather than existential risk. Regulators differ too: EU tech chief Henna Virkkunen told Reuters Friday that the bloc's AI Act is sufficient to address risks from rogue AI systems.

Context

Governments are moving in different directions. On Sunday, the Trump administration lifted its block on U.S. labs giving Britain's AI Security Institute early access to models for safety testing, according to Reuters. In Canada, Artificial Intelligence Minister Evan Solomon said in September that he had discussed a proposed "technology stability board" to evaluate AI models with Group of Seven counterparts, The Canadian Press reported. The news agency also noted that a San Francisco-based lab monitoring AI activity reported Sept. 30 that the Library and Archives Canada website was among the targets of a failed hacking attempt by an AI agent, research we covered at the time.

What to Watch

  • Publication of the letter's full text and signatory list, and responses from the labs whose incidents it cites.
  • Whether the White House, the G7 or other governments take up containment standards or an incident registry.
  • Whether the European Commission holds to its view that the AI Act already covers rogue AI as incidents accumulate.