How 3 Indian Researchers Used Claude To Find OpenAI Security Flaws | Explained

Three Indian cybersecurity researchers from startup Hacktron AI managed to uncover a chain of vulnerabilities affecting OpenAI systems and were paid a $6,500 bug bounty for responsibly reporting them. Harsh Jaiswal, Rahul Maini and Mohan Pedhapati used Anthropic’s Claude AI chatbot during their authorised security research. Their investigation began with OpenAI’s public community forum and eventually led to access involving authentication tokens, employee ChatGPT and Codex accounts, and a private software repository. The researchers stopped after demonstrating the access and reported their findings to OpenAI. The company subsequently fixed the vulnerabilities and disabled affected tokens and sessions, according to the account.