Researchers Uncover Prompt Adware Hidden in OpenAI Codex Jailbreak Toolkit

Check Point Research has detailed a jailbreak bundle for OpenAI’s Codex, called Haiou, that persistently modifies the agent’s configuration to inject advertisements for a third-party LLM gateway into its responses — a behaviour the researchers dub ‘prompt adware.’ Distributed on GitHub as an enhancement package, the bundle triggers ads on common greetings and has been forked dozens of times. Check Point warns the technique shows how third-party skills and configurations can quietly alter AI agent behaviour, a supply-chain risk as agents enter enterprise environments.