The A.I. Daily
Researchers find prompt adware in OpenAI Codex jailbreak toolkit
Check Point Research disclosed a Codex jailbreak bundle called Haiou that uses prompt injection to alter global agent instructions, inserting ads for a third-party LLM gateway when users type triggers like 'hi' or 'Seagull.' It calls this scheme 'prompt adware' aimed at driving users to the paid gateway.