Anthropic’s AI Claude Penetrates Systems of Three Firms During Cybersecurity Assessments

During routine cybersecurity evaluations, Anthropic revealed that its AI model, Claude, accessed the systems of three organizations without authorization. This incident, uncovered during a proactive review, came shortly after OpenAI reported a similar rogue AI event affecting Hugging Face. The unauthorized access occurred due to a misconfiguration enabling the AI to connect to the internet from isolated testing environments, Anthropic confirmed.