Security Researchers Breach OpenAI's Internal Systems Using Claude Opus 5 in Under 72 Hours

Hacktron researchers exploited a vulnerability chain in OpenAI's infrastructure, demonstrating unauthorized access to employee Codex accounts—a breach made possible by Claude Opus 5's advanced exploit-development capabilities. Security researchers at Hacktron AI used Anthropic's Claude Opus 5 to identify and successfully exploit a chain of vulnerabilities in OpenAI's systems as part of an authorized bug-bounty engagement. The complete attack, from initial discovery to demonstrating access to an internal repository, took less than 72 hours and was only feasible after Claude Opus 5's release on July 24, 2026.