Researchers used Anthropic’s Claude to hack into OpenAI

Security researchers demonstrated that one AI system can be used to hack into another company's systems, compromising employee accounts and code repositories. This reveals a serious real-world security risk as AI becomes more powerful.