About 400 Scientists Call for the Creation of an International Oversight Body for Advanced AI

Around 400 scientists from 34 countries have advocated for the creation of an international body to oversee the safety of advanced artificial intelligence systems. An open letter with this initiative is cited by Reuters. Among the signatories are industry pioneers Geoffrey Hinton and Yoshua Bengio. It is proposed that the new structure will work on developing industry-wide standards for model isolation and creating a registry of significant security incidents. According to the researchers' plan, this registry should eventually become public.

The authors of the appeal insist on strengthening isolation. In their opinion, advanced models with cyber capabilities should initially be completely disconnected from external networks using air gap technology. The scientists also demand independent oversight and emphasize that voluntary commitments from tech companies are not enough. Organizations that develop or test such models should immediately implement highly reliable isolation standards while governments are still preparing a more formal control system. The letter was a response to a series of incidents that exposed weaknesses in the testing and isolation of advanced systems. In particular, it refers to cases where models bypassed insufficiently protected sandboxes and compromised external systems.

Meanwhile, politicians continue to debate methods of AI regulation. EU technology chief Henna Virkkunen told Reuters on Friday that the EU's Artificial Intelligence Act is sufficient to address the risks associated with uncontrolled systems. The administration of US President Donald Trump, in turn, lifted a block that prevented American labs from providing the British Institute for AI Safety with early access to models for safety checks. Earlier, media reported that OpenAI and Anthropic have begun preparing for a potential AI catastrophe and its consequences.