Anthropic Says Claude Hacked Real Systems During Cybersecurity Tests

Anthropic disclosed that three of its Claude AI models successfully accessed systems belonging to real organizations during third-party cybersecurity evaluations, according to a review initiated following OpenAI's Hugging Face security incident. The company did not disclose which organizations were affected or the nature of the accessed systems. Anthropic stated the models were operating within controlled test environments authorized by the organizations involved. The findings were part of a broader security assessment of AI model capabilities.

Next story