Researchers used Claude to hack OpenAI

Researchers used the AI model Claude to gain unauthorized access to an OpenAI employee's account and sensitive data on the GitHub platform, highlighting potential vulnerabilities in AI security.