Check Point Research has disclosed details of a jailbreak bundle named Haiou, which targets OpenAI Codex and is designed to modify an AI agent's trusted instructions to insert advertisements for a third-party AI gateway. The researchers have categorized this phenomenon as 'prompt adware'. The disclosure indicates that the modification of trusted instructions could lead to user deception and unintended promotion of third-party services, though specific exploitation mechanisms and the full scope of affected systems remain under investigation.

What should defenders do?

Mitigation strategies likely involve restricting unauthorized modifications to AI agent configurations, implementing strict input validation for system prompts, and monitoring for unexpected changes in AI output behavior. Users and organizations utilizing OpenAI Codex should review access controls and stay informed about official OpenAI security advisories.

mediumSecurity Research

Researchers Report Prompt Adware in OpenAI Codex Jailbreak Toolkit

Check Point Research identifies Haiou jailbreak that modifies AI agent instructions to insert third-party advertisements

Cyber.TechVoidPublished 16 minutes agoUpdated less than a minute ago1 min read

Researchers Report Prompt Adware in OpenAI Codex Jailbreak Toolkit

Key Takeaways

Check Point Research has identified a jailbreak bundle named Haiou targeting OpenAI Codex.
The bundle is designed to modify trusted AI agent instructions to insert third-party advertisements.
The activity has been categorized as 'prompt adware'.
The modification raises risks of user deception and unintended service promotion.

Related Security News

highThreat Intelligence

Chinese Hackers Deploy ARTEX AI Framework in Automated Attacks

According to a report by BleepingComputer, Chinese state-aligned hackers have been using ARTEX, an open-source multi-agent AI penetration-testing framework, to conduct automated attacks against more than 200 organizations across 14 countries. The campaign, reported in November 2026, leverages the AI framework to automate penetration testing and exploit vulnerabilities in target environments. Specific exploitation methods and targeted sectors are not detailed in the available summary. The use of open-source AI frameworks in cyber operations is identified as a growing trend in threat intelligence.

BleepingComputer2 days ago1 min read