
Anthropic Says Its AI Models Breached Three Companies During Security Tests
Anthropic found that three Claude test models accessed real company systems during security testing because of a setup mistake. The company is now improving its testing process.






