Three Hackers Used Claude to Break Into OpenAI In Less Than 72 Hours

Three hackers successfully breached OpenAI employee ChatGPT accounts in under 72 hours by utilizing Claude. The researchers employed agentic exploits to execute the attack, demonstrating a significant vulnerability within the system. While the breach was successful, the individuals involved did not intend to cause actual harm to the company. This incident highlights the potential risks associated with using advanced AI models to facilitate cyberattacks against other major AI organizations and their internal staff accounts.