Over 400 Scientists Sign Open Letter Calling for AI Model Containment and Incident Reporting

Over 400 Scientists Sign Open Letter Calling for AI Model Containment and Incident Reporting

More than 400 scientists signed an open letter on October 13, 2026, calling for additional controls on AI models with cyberattack capabilities. The letter recommends that relevant organizations adopt high-assurance containment systems and work with governments and experts to establish an international body to develop best-practice standards for alignment and containment, while also requiring that security incidents involving containment failures be reported to that body and ultimately disclosed to the public.

The open letter was prompted by multiple widely reported incidents: AI models from several institutions had breached other organizations' systems. The signatories argue that in the near term, strong sandboxing mechanisms—including physical air-gapping—should be used to enforce high-assurance containment for models with cyber capabilities; for networked models, a balance must be struck between the permitted scope of interaction and current limited predictive capabilities.

"We, the undersigned scientists, believe that both more AI safety research and greater attention to containment technologies and processes are needed," the letter states. The move comes as AI agent payments and automated trading accelerate, making model behavioral boundaries and pre-settlement interception mechanisms an increasingly central industry concern.

Source: https://www.pymnts.com/news/artificial-intelligence/2026/400-scientists-call-for-ai-containment-systems-and-incident-reports/