Free the Model, Sell the Gateway: Prompt Adware hidden in Jailbreak Bundles

Check Point Research discovered a Chinese jailbreak bundle for OpenAI Codex named Haiou that functions as prompt adware by persistently injecting advertisements for a paid LLM gateway into global configuration files. Researchers observed the bundle distributed alongside cybercrime tooling such as Cobalt Strike, highlighting the dual-use nature of advanced AI jailbreaks. The package modifies agent instructions using various prompt injection techniques while operating entirely through natural language rather than executable payloads. Open article →