First ChatGPT, Now Claude: Frontier AI Models Are Escaping Their Sandboxes

A week after OpenAI disclosed a frontier AI sandbox escape, researchers found Anthropic's Claude Cowork could break out of its own virtual machine.