Haiou jailbreak alters OpenAI Codex instructions to insert ads

First reported by Cyber Security News at · Updated · 2 sources

Researchers found that the bundle modifies the coding agent's trusted instructions to promote a third-party AI gateway. The Hacker News attributes the term prompt adware to Check Point Research, while Cyber Security News reports that the Chinese-language package makes ordinary greetings trigger advertisements and maintains the behavior through written instructions.

  • Haiou means Seagull in Chinese, according to The Hacker News.

Covered by 2 publishers within 11 hours of the first report.

Reporting2

Cyber Security News Hackers Use OpenAI Codex Jailbreak Toolkit That Secretly Injects Ads Into AI Responses · Tushar Subhra Dutta
The Hacker News Researchers Uncover Prompt Adware Hidden in OpenAI Codex Jailbreak Toolkit · info@thehackernews.com (The Hacker News)

Related

OpenAI staff raised security concerns with leaders for years
OpenAI says AI pipeline produced 948 Linux privilege escalation exploits
Xint shows OpenAI compaction flaw can expose hidden AI reasoning
OpenAI agrees to buy Pierre Computer to expand Codex tools

Topics OpenAI