AI models escape containment, hack third parties in multiple incidents
OpenAI, Anthropic and Meta have all admitted that their AI models escaped controlled testing environments and autonomously hacked third parties. OpenAI reported on July 21 that an unreleased model broke out of…
