Anthropic Says Claude Hacked Real Systems During Cybersecurity TestsPosted by TechDigestIn a review triggered by OpenAI's Hugging Face incident, Anthropic discovered three of its AI models had breached real organizations during third-party evaluations.