More than 400 signatories, including Nobel laureate Geoffrey Hinton and Turing Award winner Ron Rivest, are urging frontier AI model makers to work with governments on an international standards body for AI containment and safety. The 415 scientists also called on companies to “exercise caution” and make cyber-capable models default to air-gapping, which physically disconnects them from unsecured networks.

The proposed body would bring together organisations, governments, AI experts and cybersecurity experts to establish best practices for AI alignment and containment in the public interest. The signatories also want it to maintain a public record of containment failures. In the letter, they said incidents involving AI systems could reflect both failures to ensure predictable model behaviour and lapses in cybersecurity practices that stronger sandboxes and procedures might prevent.

The letter warned that without immediate changes to AI development and deployment practices, such incidents would continue and could cause greater harm to exposed critical infrastructure. The article also cited a Eurobarometer survey published on 14 October: 42% of European small and medium-sized enterprises and mid-cap companies already use AI, while another 21% plan to do so. The article recalled earlier safety calls, an open letter signed by 120 companies in August seeking more collective cyber-defence action, and separate calls by OpenAI and Anthropic leaders to pace frontier AI development.

More than 400 signatories, including Nobel laureate Geoffrey Hinton and Turing Award winner Ron Rivest, are urging frontier AI model makers to work with governments on an international standards body for AI containment and safety. The 415 scientists also called on companies to “exercise caution” and make cyber-capable models default to air-gapping, which physically disconnects them from unsecured networks.

The proposed body would bring together organisations, governments, AI experts and cybersecurity experts to establish best practices for AI alignment and containment in the public interest. The signatories also want it to maintain a public record of containment failures. In the letter, they said incidents involving AI systems could reflect both failures to ensure predictable model behaviour and lapses in cybersecurity practices that stronger sandboxes and procedures might prevent.

The letter warned that without immediate changes to AI development and deployment practices, such incidents would continue and could cause greater harm to exposed critical infrastructure. The article also cited a Eurobarometer survey published on 14 October: 42% of European small and medium-sized enterprises and mid-cap companies already use AI, while another 21% plan to do so. The article recalled earlier safety calls, an open letter signed by 120 companies in August seeking more collective cyber-defence action, and separate calls by OpenAI and Anthropic leaders to pace frontier AI development.