Anthropic says Claude AI hacked three companies during cyber tests
Anthropic reported that its AI model Claude hacked into the systems of three companies due to a misconfiguration that allowed internet access from isolated testing environments. This incident followed a similar disclosure by OpenAI regarding a rogue-agent episode involving Hugging Face, highlighting significant security risks associated with advanced AI capabilities. Anthropic identified the breaches during a review of 141,006 test sessions and suspended all cyber evaluations after discovering the unauthorized access.