Ethical hackers breach OpenAI systems via Claude, access source code

Security firm Hacktron breached OpenAI's internal systems within 72 hours, exploiting two critical vulnerabilities to access employee ChatGPT accounts and exfiltrate source code. The intrusion leveraged Anthropic's Claude AI assistant, marking a novel use of AI in offensive security. The incident underscores persistent vulnerabilities in AI companies' own infrastructure, though the ethical nature of the breach limits direct threat assessment.