AI-Built Exploit and Sign-In Flaw Opened Path to Internal OpenAI Code
Source SecurityWeek
Rated by Threat Radar
Data as of 18 Sep 2026, 12:46 UTC
What to do
AI-suggested
OpenAI fixed the sign-in token permission flaw by narrowing token permissions and revoking affected tokens and sessions approximately 14 hours after disclosure. Discourse patched the libheif vulnerability within two days and implemented image-processing sandboxing as an additional defense layer. Users and employees should ensure their tokens and sessions are revoked and updated. Since fixes are available and deployed, no further immediate action is required beyond applying these patches and monitoring for unusual account activity.
Description
Researchers at Hacktron demonstrated a chained exploit involving an unpatched vulnerability in the libheif library used by OpenAI's community forum (Discourse) and a flaw in OpenAI's sign-in token permissions. This allowed remote code execution on the forum and subsequent takeover of OpenAI employee ChatGPT and Codex accounts, leading to access to internal code repositories. The libheif flaw was fixed upstream but not flagged as a security issue, and OpenAI quickly fixed the token permission issue after disclosure. The exploit could have allowed access to connected services like GitHub and Slack, though no evidence of such access was confirmed.