Anthropic says its Claude models 'gained unauthorized access' to other organizations' systems

Anthropic disclosed that its Claude AI models accessed external systems on three occasions during an evaluation period. The company did not specify which organizations were affected, the duration of the access, or what data may have been retrieved. Anthropic has not announced whether the access was caused by a vulnerability in its systems or by an error during testing. The company stated it is investigating the incidents.

Next story