Researchers used Anthropic’s Claude to hack into OpenAI

7.7 relevance

Novel use of AI agents for cybersecurity, highly relevant to AI/ML and security.

AI/ML techcrunch.com

Answer the board on AI readiness

Arrive with evidence, not opinion. See how your foundation benchmarks against your peers, get a clear read, and leave with a ranked place to start.

Why It Matters

For a platform engineer focused on AI security and supply chain risk, this incident demonstrates that even well-funded AI companies are susceptible to chained exploits via unpatched third-party dependencies, and that frontier models like Claude Opus 5 can autonomously weaponize such vulnerabilities—making AI-assisted red teaming both a threat and a necessary defense.

Key Takeaways

Audit all third-party dependencies for unpatched vulnerabilities, especially those without CVE tracking, and integrate AI-driven penetration testing into your security pipeline to stay ahead of attackers using the same tools.

Researchers used Anthropic’s Claude to hack into OpenAI

Summary

Security researchers at Hacktron AI used Anthropic's Claude Opus 5 to chain two vulnerabilities—a libheif memory bug in Discourse forum software and a subsequent account takeover—to access OpenAI employee ChatGPT and Codex accounts, earning a $6,500 bug bounty. The exploit succeeded only after Opus 5 was released, as Opus 4.8 failed to produce a working exploit, highlighting how rapidly improving AI models can lower the barrier for sophisticated attacks. The libheif bug had been patched months earlier but lacked a CVE identifier, leaving Discourse running the vulnerable version.

Author

Aditya Mehta, Rebecca Bellan