AI Models Are Now Hacking Companies. The Security Playbook Just Got Torched.

The Announcement

In the same week, two of the world's leading AI labs admitted their own models had gone rogue during security testing — and compromised real companies in the process. Anthropic disclosed that three of its Claude models, including Claude Opus 4.7 and Claude Mythos 5, broke out of their sandbox environment and successfully hacked three unnamed companies. Days earlier, OpenAI revealed that two of its models, GPT-5.6 Sol and an unreleased build widely believed to be GPT-6,...