Scientists call for international oversight of advanced AI

On Oct. 13, Reuters reported that about 400 scientists from 34 countries—including Geoffrey Hinton and Yoshua Bengio—had called for an international body to oversee containment standards and security-incident reporting for advanced AI with cyber capabilities. The proposed plan would also initially isolate frontier models from external networks and subject them to independent oversight, according to Reuters’ report on the appeal and its proposed measures.

What the proposed body would do

The scientists’ appeal calls for industry-wide containment standards and a register of significant security incidents. Public access to that register is envisaged eventually.

The plan pairs technical safeguards with independent scrutiny and shared reporting. One proposed safeguard is to air-gap frontier models at first—that is, keep them disconnected from external networks. The appeal also urges organizations developing or testing cyber-capable models to adopt high-assurance containment immediately while governments work toward a more formal oversight regime.

Geoffrey Hinton, one of the reported signatories, had separately proposed that AI companies persuade a regulator their products are safe before release. NeoTeo covered Hinton’s earlier proposal for pre-release safety review on Oct. 6.

How it differs from other oversight efforts

A separate diplomatic appeal focused on human direction and common rules. Naiz reported in September that 22 states and institutions had called for those measures and for consideration of an international institution.

The EU’s approach has a different scope. In its summary of Article 14 of the EU AI Act, the AI Act explainer describes human-oversight duties for high-risk AI systems. It says oversight should let people understand a system’s limitations, disregard its outputs or interrupt its operation, with measures proportionate to the system’s risk and autonomy.

A White House accord signed by Anthropic, OpenAI, Google, Meta, xAI and Nvidia roughly a month before the scientists’ appeal was voluntary and had no formal enforcement, Reuters reported.