Anthropic says Claude models breached three organizations after escaping a misconfigured cyber evaluation environment run with Irregular.

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

Frontier AI systems are increasingly capable of translating narrowly defined objectives into complex, real-world cyber ...

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 a review triggered by OpenAI’s recent disclosure found three real-world intrusions caused by a misconfigured AI testing environment.

Anthropic has admitted that its Claude AI accidentally hacked three real organisations during cybersecurity testing after a ...

Anthropic revealed that its Claude AI models accessed real systems during cybersecurity tests due to a misconfigured ...

Anthropic has disclosed three incidents in which its Claude models accessed real-world systems during cybersecurity tests, ...

The company says the incidents were not the result of the models deliberately attempting to escape their testing environment but stemmed from an evaluation setup that mistakenly allowed access to the ...

Anthropic says 3 Claude models breached real organizations after misconfigured CTF evaluations exposed them to the open internet and production system ...

Anthropic says a review found Claude models accessed real-world systems after a third-party AI cybersecurity evaluation ...

Anthropic says three Claude models hacked real company systems during safety tests after a misconfiguration gave them ...