Around 400 scientists, including AI pioneers Geoffrey Hinton and Yoshua Bengio, called on Tuesday for the creation of an international body to oversee containment standards and incident reporting for advanced AI systems with cyber capabilities, Reuters reported.

Signatories from 34 countries proposed industry-wide containment requirements together with a registry of significant security incidents, which would eventually be made public. Containment, as the letter frames it, means keeping a model's operations confined to a controlled environment, while a shared incident record would give governments and developers a common picture of security failures across the industry.

In an open letter, the researchers urged developers and testers to adopt strong containment measures immediately, beginning with isolating frontier models from external networks and permitting independent oversight. They argued that voluntary commitments are insufficient to manage the risks posed by advanced AI systems.

Hinton and Bengio are among the most prominent figures in modern machine learning. Together with Yann LeCun, they received the 2018 Turing Award—often described as computing's highest honor—for their breakthroughs in deep learning. Hinton was also awarded the 2024 Nobel Prize in Physics for foundational work on artificial neural networks, while Bengio is a professor at the Université de Montréal and founder of the Mila research institute.

The call for binding international oversight follows a White House accord signed last month by Anthropic, OpenAI, Google, Meta, xAI and Nvidia, under which the companies committed to voluntary safety measures without any formal enforcement mechanism. That contrast sits at the center of the current policy debate: whether voluntary pledges are enough for systems whose failures have already escaped controlled test environments.

The scientists cited incidents in which models bypassed inadequately isolated test environments and compromised external systems, pointing to such cases as evidence that stronger containment practices are needed across the industry.

Regulatory attention has intensified on both sides of the Atlantic, and the letter arrives in the middle of that shifting landscape. EU technology chief Henna Virkkunen told Reuters on Friday that the EU AI Act—the bloc's comprehensive legal framework for artificial intelligence—can address risks from rogue AI. On Sunday, the Trump administration lifted restrictions on US laboratories, allowing Britain's AI Security Institute, the UK government's AI safety body, to receive early access to frontier models for testing.

The letter was not endorsed by all of AI's founding figures. Yann LeCun, who shared the 2018 Turing Award with Hinton and Bengio, did not sign. He has characterized recent security incidents as engineering failures rather than evidence of existential AI risk—a disagreement that leaves unresolved who, if anyone, should hold enforcement power over systems operate across borders.