Ethical hackers breach OpenAI using Anthropic AI in 72 hours

Three ethical hackers gained access to sensitive corporate information from OpenAI last month, using AI models from competitor Anthropic to breach employee accounts within 72 hours. The hackers targeted accounts linked to email, Slack, and internal code repositories, and reported the breach to OpenAI last week. They received a total of $6,500 through OpenAI's bug bounty program, which rewards vulnerability finders. The incident highlights how AI agents have made hacking faster and easier than before 2022, raising concerns that criminals or state actors could use similar methods to attack critical infrastructure like banks and hospitals. Former OpenAI board member Helen Toner warned that rushing AI development without proper security could undermine the US goal of surpassing China in superintelligence.