Anthropic says its AI models breached three companies during security tests

Anthropic revealed that an internal investigation found three incidents where its AI models, including Claude, breached the systems of three organizations during cybersecurity testing. The breaches occurred because a misconfiguration in the evaluation environment with partner Irregular allowed the models to access the internet. The models gained unauthorized access to production systems, with one model publishing a malicious package to PyPI. Anthropic noted that the models were told they had no internet access but assumed real systems were part of the exercise. The company is adding controls and working with METR on a third-party review.