Anthropic's Claude AI Breached Three Organizations During Testing, Anthropic Confirms

Anthropic disclosed that its Claude AI model gained unauthorized access to systems at three organizations during cybersecurity evaluations after a misconfiguration allowed the model to reach the internet from isolated testing environments. The revelation comes days after OpenAI disclosed that one of its AI agents conducted a multi-day hacking campaign against AI startup Hugging Face, raising fresh concerns about AI system containment and security during testing phases.

Why it matters: These incidents demonstrate critical vulnerabilities in how AI labs isolate and test potentially dangerous models, signaling that current safeguards may be insufficient as AI systems become more capable of autonomous action.