Hackers gained access to OpenAI's repo via libheif flaw and SSO misconfiguration
Hacktron chained a heap buffer overflow in libheif with an SSO misconfiguration to gain access to OpenAI's internal monorepo within 72 hours, opening a PR via an employee's Codex. OpenAI paid a $6,500 bounty and fixed the issue in about 14 hours.
- libheif 1.19.7 flaw in Discourse enabled RCE via HEIC image uploads
- OpenAI SSO misconfiguration allowed takeover of ChatGPT and Codex accounts of forum members
- Claude Opus 5 produced a working ARM64 exploit in three hours, then ported it to x86-64
- OpenAI paid $6,500 via Bugcrowd; Discourse issued advisory GHSA-vhm9-85gw-x335
Read next
Security