Latest / Frontier models and safety
Haiou jailbreak bundle for Codex injects ads into agent instruction files
See how this story connectsCheck Point Research analysed the Haiou jailbreak bundle, a prompt package targeting OpenAI Codex, which it classifies as prompt adware. The bundle installs security-focused Skills and modifies Codex configuration files so that advertisements for an LLM gateway website are injected into later agent responses. Check Point found the bundle on a misconfigured webserver of a threat actor that ran Cobalt Strike, alongside Ladon and VulnClaw, and says this does not show the bundle was designed for cybercrime.
Why it matters
Anyone who installs the bundle has Codex configuration files altered so that ads for a third-party LLM gateway appear in later agent responses.
- When
- 3 Nov 2026
- ID
WTA-N-2026-0701- Where
- International
- Who
- Check Point Research, OpenAI
- Topics
- Cybersecurity, AI agents, AI incidents
Summary written with AI by wheresthe.ai, from Check Point Research and the reporting credited above. Checked against the source on 4 Nov 2026.
Source copy: saved on 4 Nov 2026 to check this story and not republished, because the text belongs to its publisher. Also held by the Internet Archive.
Cite this story
WTA-N-2026-0701: “Haiou jailbreak bundle for Codex injects ads into agent instruction files.” wheresthe.ai, 3 Nov 2026. https://wheresthe.ai/d/intl-haiou-jailbreak-bundle-for-codex-injects-ads-into-agent-2026-11/ (source: Check Point Research).
Spotted an error? Email [email protected].