400 Researchers Including Bengio and Hinton Urge AI Labs to Airgap Cyber-Capable Models
Summary
More than 400 researchers, including Yoshua Bengio and Geoffrey Hinton, published an open letter Tuesday urging AI labs to adopt high-assurance containment systems and airgap cyber-capable models. The letter says failures like OpenAI's July breach of Hugging Face were preventable with stronger sandboxes. It also calls for an international standards body to track and publish containment failures.
Key Points
- Anthropic cofounder Jack Clark says in a Monday Substack newsletter that full RSI becomes possible in 2027.
- Reuters reports Anthropic and OpenAI seek influence over what third-party evaluators can publish.
- The SPUR coalition adds The New York Times, The Atlantic and The Economist, reaching 80 members globally.