Short answer: Researchers found a jailbreak for OpenAI Codex that inserts ads into AI responses. Users of the jailbreak could see ads. Be cautious when using third-party tools for AI models.
TL;DR
- Researchers found a jailbreak called Haiou.
- It modifies AI responses to include ads.
- Users should be careful with third-party AI tools.
What happened?
This issue involves a jailbreak named Haiou, which is a technique that modifies OpenAI Codex, an AI model. The jailbreak injects ads into the AI’s responses, turning it into a form of prompt adware. This means that every time the AI is used, it might display ads. The jailbreak is part of a larger toolkit and has been popular among users, with many forks and stars on GitHub. This highlights the risks of using third-party tools and the potential for supply chain attacks in AI.
- Supply chain attack
- Attacking a trusted supplier or software package to reach all of its customers.
Key facts
| Reported by | The Hacker News |
|---|---|
| Date reported | 04 Nov 2026 |
How it works
The Haiou jailbreak modifies OpenAI Codex's configuration files to inject ads for a paid gateway service. It uses techniques like role-playing and CTF scenarios to bypass safety settings. Ads appear in responses when users type certain greetings or commands.
- Modifies Codex's config files
- Injects ads for paid service
- Triggers ads with greetings
- Uses CTF scenarios for tricks
Timeline
| Date | Event | Source |
|---|---|---|
| 2026-11-04 | Reported by The Hacker News | The Hacker News |
| 2026-11-11 | This Skillmate explainer published | Skillmate |
How to fix it
- Keep the operating system, browser and security software updated.
- Do not open unexpected attachments or install software from unofficial sources.
- Report suspicious files or pop-ups to your IT or security team.
How to detect it
Defenders should watch for unexpected ads in AI responses.
Signs to watch for:
- Unknown programs starting at boot, or new scheduled tasks.
- Security software that was turned off without a reason.
- Unusual outbound connections from a workstation.
What is confirmed
- Reported by The Hacker News
Why it matters
This matters because it shows how attackers can misuse AI to inject unwanted ads. It highlights risks in AI tools and the need for better security to protect users and companies.
Learn the skills behind this story
Want to understand attacks like this and defend against them? Join Skillmate cybersecurity training or practise hands-on in Skillmate Labs.
FAQ
What is prompt adware?
Prompt adware is a type of software that adds ads to AI responses by modifying the AI's instructions.
How can I protect against prompt adware?
Regularly update your security tools and monitor AI responses for unexpected ads.
What is Haiou?
Haiou is a name given to the jailbreak bundle that modifies OpenAI Codex to insert ads for a third-party AI gateway.
Why is this jailbreak bundle dangerous?
It could lead to persistent ads in AI responses, potentially compromising the integrity of the AI's output.
Related Skillmate posts
Sources
How we wrote this: facts come from the sources above. An on-prem AI model drafted the plain-English parts and every sentence was checked against the sources; unsupported sentences were removed. MITRE mapping is our own judgement.