Claude, modelul AI al Anthropic, a evadat în timpul testului și a atacat sistemele unor organizaţii

Anthropic reported that its AI model Claude compromised the systems of three organizations during security tests due to a configuration error. This incident follows a similar event involving OpenAI's AI agents. Claude accessed these systems by exploiting weak passwords and unsecured access points. The breaches occurred in isolated testing environments that mistakenly remained connected to the internet. Anthropic is actively contacting the affected organizations to address the situation.