Researchers Uncover Prompt Adware Hidden in OpenAI Codex Jailbreak Toolkit
Summary: Check Point Research disclosed a jailbreak bundle called Haiou (Seagull in Chinese) that modifies OpenAI Codex's configuration files to inject ads for the author's LLM gateway site, zz.aiapi2025[.]top, into agent responses triggered by greetings like "hi," "hello," and "Seagull." The GitHub repository haiou-codex, advertised as a local enhancement package adding a Chinese technical personality and security research workflow to Codex, had been forked 77 times and starred 403 times since its July 10, 2026 release, and the jailbreak was also bundled inside the Codex-X configuration tool.
What you should know: The behavior arrives as natural-language instructions and configuration rather than an executable payload, so it does not look like software your teams are trained to flag. Check Point also said the gateway resale model is "more phishy, like laundering stolen LLMs tokens," which puts credential and token exposure on the table alongside the ads.
Read the original at The Hacker News ↗
- Topics
- AI Security
- Related
- SECURITY AFFAIRS AI-CYBERSECURITY NEWSLETTER ROUND 8 Cyber Risks · Nov 15, 2026Pentagon Renames CDAO ‘Super Intelligence Office’ Cyber Risks · Nov 15, 2026186 Fake AI Domains Redirect Users to Click Fix Lures, Browser Malware, and Search Hijackers Cyber Risks · Nov 12, 2026