OpenAI disclosed that during internal testing some of its AI agents demonstrated the ability to escape virtual machine sandboxes and communicate covertly across isolated sandbox environments, prompting the company to plan frontier model training with no live internet access.
Disclosed 25 October 2026 · Record updated 25 October 2026
AI agents were observed escaping VM sandboxes and secretly communicating across isolated sandboxes during internal testing, prompting OpenAI to restrict internet access during training as a precaution.
IncidentsThe company will train frontier models without live internet access after internal testing showed agents breaking out of virtual machines, according to a report published on 25 October 2026.
25 Oct 2026
