AI-Built Exploit and Sign-In Flaw Opened Path to Internal OpenAI Code

Summary

Researchers at Hacktron used Claude to build a working exploit for a libheif flaw reached through ImageMagick, which Discourse passed HEIC/HEIF uploads to on OpenAI's community forum, community.openai.com. Chained with a separate OpenAI sign-in weakness, where forum sign-in tokens carried excessive permissions and granted full API access to linked ChatGPT and Codex accounts, the team took over an employee account and opened a pull request in an internal repository. OpenAI says the review found limited reads of private-repository metadata and commits, and it narrowed the token permissions and revoked affected tokens and sessions.

Mitigation

OpenAI narrowed the permissions on Community sign-in tokens and revoked affected tokens and sessions. Discourse had a fix ready within two days and added image-processing sandboxing as an extra layer of defense, then published a security advisory.

Related items

• MediumGHSA-68g8-gh8f-7rhg: pgEdge RAG Server discloses part of the configured provider credential through API error messagesSame vendor · GitHub Advisory Database
• InfoHow Amplitude ships product improvements in 1–2 days with OpenAISame vendor · OpenAI Blog
• InfoEndor Labs + OpenAI: Strengthening the Defense Factory With Accelerated Remediation | Endor LabsSame vendor · Endor Labs Blog
• InfoJensen Huang built an AI kingdom. Now he’s gone, will Nvidia’s ramparts be breached?Same vendor · The Guardian Technology
• LowAI chatbots are recommending fake retail websites, study findsSame vendor · The Verge (AI)