Researchers use Claude Opus 5 to chain flaws and access OpenAI staff accounts

Security researchers used Anthropic’s Claude Opus 5 to develop a working exploit chaining a forum image-processing vulnerability with OpenAI’s single sign-on system, enabling takeover of OpenAI employee ChatGPT and Codex accounts and limited internal repository access. The work was disclosed under a bug bounty; OpenAI paid $6,500 and patched the issue. The episode illustrates how advanced models can accelerate sophisticated offensive security research under human direction.