Security researchers used Claude to help them hack into OpenAI
Three Hacktron researchers used Claude Opus to breach OpenAI employee accounts through a Discourse HEIF flaw, reaching the Monorepo within 72 hours.
A three-person team at Hacktron used Anthropic's Claude Opus 4.8 and 5 to exploit a HEIF image-processing flaw in Discourse, achieving RCE on Discourse Cloud and access to OpenAI's community forum instance within roughly a day of Claude Opus 5's July 24 launch. They reached OpenAI's GitHub Monorepo through employee accounts and proved access with a pull request from an employee's Codex account. Their HEIF Heist tooling adapted to targets including OpenAI, Slack, Meta, and GitHub Enterprise for under $3,000 in tokens, and was detected by only one target, Shopify. Discourse and OpenAI have since fixed the reported vulnerabilities, and OpenAI paid a $6,500 bounty.
- Three researchers achieved Discourse Cloud RCE within a day using Claude Opus 5
- Accessed OpenAI GitHub Monorepo and sent PR from an employee's Codex account
- Tooling adapted to OpenAI, Slack, Meta, GitHub Ent for under $3,000 in tokens
- Only Shopify detected the activity; OpenAI paid a $6,500 bounty
- Discourse and OpenAI have since fixed the vulnerabilities
- Threat actors
- Hacktron
- Victims
- OpenAI
- Organizations
- HacktronWall Street Journal
- Countries
- United States
- AI models
- Claude Opus 4.8Claude Opus 5
From the original article
1 min read at theverge.com
A team of three independent security researchers at Hacktron says it took less than 72 hours for them to hack into OpenAI employee accounts using Anthropic's Claude Opus 4. 8 and 5, the Wall Street Journal reports. They were able to access OpenAI's GitHub repository, called "Monorepo," which reportedly contains "OpenAI's algorithmic secrets," according to the Wall Street Journal's sources.
By Stevie Bonifield · theverge.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.