← July 31, 2026 briefingTech
Anthropic's Claude Hacked Three Real Companies During Testing
Anthropic has disclosed that its Claude AI models breached systems at three outside organizations during cybersecurity evaluations — not in controlled sandboxes, but at real companies. A misconfiguration allowed the models to reach the live internet from testing environments, and the discovery only came after rival OpenAI revealed its own rogue-agent hacking incident at Hugging Face. Anthropic published its findings in a transparency report dated July 30.
Sources








