Anthropic said the OpenAI event spurred its engineers to review similar cybersecurity evaluations by Claude models. The audit ...

Anthropic's Claude AI models breached three companies' live systems during cybersecurity tests, with the victims unaware ...

One of Anthropic's Claude models built and uploaded a malicious Python package to PyPI during a botched security evaluation, where it ran on 15 real systems and stole credentials from a security ...

Anthropic says three Claude AI models accessed live company systems during misconfigured cybersecurity tests, exposing ...

Anthropic says Claude models escaped security tests, published a malicious PyPI package, and accessed real production systems.

Researchers found AI coding agents build less reliable pipelines when forced into structured formats — DataFlow-Harness ...

Anthropic says three Claude models escaped sealed test environments and breached three real organizations after a ...

Wrote and published malware during tests, which is apparently OK because leaky test environments were the real problem ...

Anthropic reviewed 141,006 of its own test runs after OpenAI's Hugging Face hack, and found three Claude models had broken ...

Anthropic found three cybersecurity evaluation incidents in which Claude models gained unauthorized access to real organizations.

According to Anthropic, the third cybersecurity incident involved an unnamed “internal research test model.” It compromised ...

ARC-AGI-3 benchmark gains its first fully open-source agent: NIMI's Tycho writes Python code as falsifiable hypotheses about ...