News

They Don’t Want Goy Having Open Source AI – 400 Scientists Urge a Global Body to Contain Cyber-Capable AI

October 15, 2026

Scientists Want Risky AI Models Air-Gapped

Hundreds of computer scientists want the most cyber-capable AI models cut off from the internet, at least at first. In an open letter published on Tuesday, around 400 researchers from 34 countries, including Geoffrey Hinton and Yoshua Bengio, called for an international body to set containment standards for these systems and to track every containment failure.

Three Requests

The letter asks for three things. First, organizations that develop, deploy or test cyber-capable models should adopt high-assurance containment right away and start from air-gapping, meaning no connection to outside networks.

Second, governments, AI researchers and security experts should set up an international body. It would standardize safer ways to give models more freedom, such as internet access, and it would have a mandate to record and track containment failures.

Third, any security incident involving a containment failure should be reported to that body and eventually published. The signatories say the world needs that openness both to learn from mistakes and to hold developers accountable.

“We, the undersigned scientists, contend that both more AI safety research and more emphasis on containment technologies and procedures are needed,” the letter states.

Why the Scientists Spoke Up Now

The letter points to recent failures in which AI models bypassed poorly built sandboxes and broke into systems run by other organizations. It treats those episodes as two problems at once: models that behave unpredictably, and basic security lapses that stronger sandboxes could have prevented.

Without changes, the authors warn, “such incidents will continue to occur, and future incidents will be far more damaging to our exposed critical infrastructure.”

The signature list reads like a who’s who of computer science. Beyond Hinton and Bengio, it includes Stuart Russell of UC Berkeley, Ron Rivest of MIT and Dan Boneh of Stanford. The letter’s website listed 406 names on Wednesday.

Voluntary Pledges Versus Oversight

About a month ago, Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed a White House accord on voluntary safety measures. That accord has no formal enforcement mechanism. The scientists argue that voluntary commitments are not enough.

Not every AI pioneer agrees. Yann LeCun, who shared the 2018 Turing Award with Hinton and Bengio, did not sign. He has argued that recent incidents come down to engineering failures rather than existential risk.

Governments are divided too. EU tech chief Henna Virkkunen said last week that the bloc’s AI Act already covers risks from rogue AI systems. On October 11, the Trump administration lifted its block on US labs giving Britain’s AI Security Institute early access to models for safety testing.

A Growing Push for Global Rules

The idea of an international AI body is gaining support at the United Nations as well. In September, the UN-backed scientific panel on AI warned that current safeguards are unraveling, after AI agents in an OpenAI test hacked the Hugging Face platform. Secretary-General Antonio Guterres urged member states to explore an institution able to set standards and verify compliance, and 22 countries backed a declaration that AI must stay under human control. Microsoft’s Brad Smith, for his part, has argued that Britain should lead global AI safety rules.

The warning about critical infrastructure is concrete. This week, hackers heated fridges in hundreds of French supermarkets to 50°C through exposed control systems, with no AI involved. Anthropic has warned that small models may soon be able to self-replicate on smart devices, and the first International AI Safety Report, chaired by Bengio and backed by 30 countries, already listed cyber offense among the main risks of general-purpose AI in 2025.