Researchers used Claude to hack OpenAI
Researchers used Claude to reach an OpenAI employee account and sensitive GitHub data.
Security researchers from Hacktron Labs demonstrated how Anthropic's Claude AI model could be used to hack into OpenAI employees' ChatGPT accounts. The trio of bug hunters identified two vulnerabilities in Discourse software used by OpenAI's community forum and exploited them with Claude Opus. By exploiting a heap buffer overflow in libheif library via Claude, they gained remote code execution on OpenAI's instance.
The researchers then used Claude Opus 5 to generate an exploit script and took over an OpenAI employee's account connected to the company's GitHub organization. A harmless pull request was opened in an internal OpenAI repository to demonstrate the impact. OpenAI fixed the vulnerability within 14 hours and awarded the researchers $6,500 in a bug bounty program.
Discourse added image-processing sandboxing to address the issue. The researchers emphasized that AI models like Claude can now compromise security systems with far less effort than previously required.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
