Security Team Used Claude to Breach OpenAI Employee Account, WSJ Reveals

ChengleiSi · x · 2026-09-18

A WSJ exclusive reveals an independent bug-hunting team disclosed a major OpenAI breach two weeks after the Hugging Face incident: using Anthropic's Claude-powered tooling, they accessed an OpenAI employee's ChatGPT account and could read and suggest changes to the company's private software cache. The team uses AI to find vulnerabilities before malicious actors and wants to work with frontier labs and internet-critical systems.

Related event: Researchers Use Claude Opus 5 to Breach OpenAI's Internal Systems for Bug Bounty(59 posts)→

Original post →

More from Models

Models channel →