Security1 hOctober 16, 2026, 13:30

Hackers breached OpenAI internal repo via malicious HEIF image and SSO flaw

Hacktron AI researchers chained a libheif heap overflow on community.openai.com with an OpenAI SSO identity flaw to gain RCE, then took over ChatGPT and Codex accounts of anyone signed into the forum, including employees. An employee's Codex opened a pull request in OpenAI's internal monorepo. OpenAI fixed the SSO issue in about 14 hours and paid $6,500.

DEV Community

Open source

Hackers breached OpenAI internal repo via malicious HEIF image and SSO flaw

Chain: HEIF image → RCE on forum → account takeover via OpenAI SSO
libheif 1.19.7/1.19.8 shipped unpatched in Debian 12 and 13
Discovery to internal repo access took under 72 hours
OpenAI paid $6,500 for the SSO finding only; forum was out of scope

#OpenAI#ChatGPT#Codex#Discourse

Zod repo backdoored: malicious commits blocked before reaching npm

Security

FakeGit malware campaign returns with 17,610 malicious GitHub repos

Security

GitHub lost 3,800 repos after malicious VSCode extension

Security

Hackers breached OpenAI internal repo via malicious HEIF image and SSO flaw

Zod repo backdoored: malicious commits blocked before reaching npm

FakeGit malware campaign returns with 17,610 malicious GitHub repos

GitHub lost 3,800 repos after malicious VSCode extension

Hackers hide malware commands in blockchains as malicious activity jumps 440%