Executive Summary
An OpenAI Codex jailbreak toolkit reportedly alters trusted instructions to inject third-party advertisements. Teams operating AI coding agents or other instruction-driven LLM applications should review their agent configuration and untrusted extension supply chain.
What Happened
Cybersecurity researchers have disclosed details of an OpenAI Codex jailbreak bundle that's designed to modify an artificial intelligence (AI) agent's trusted instructions to insert advertisements for a third-party AI gateway. The phenomenon has been codenamed prompt adware by Check Point Research. Called Haiou (meaning Seagull in Chinese), the jailbreak modifies OpenAI Codex's configuration
Affected Products
- OpenAI Codex
- Haiou
Vulnerability / Attack Technique
Jailbreak-based prompt injection
Technical Mechanism
The Haiou jailbreak modifies OpenAI Codex’s configuration and the AI agent’s trusted instructions. This causes advertisements for a third-party AI gateway to be inserted into the agent’s behavior, according to the article summary.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Relevant Frameworks
RealGround Analysis
This matters because trusted instructions are part of an agent’s control boundary, and modifying them can redirect agent behavior without changing the underlying model. Teams should monitor configuration integrity, evaluate jailbreaks and extensions as potential supply-chain inputs, and test whether agent outputs or actions can be influenced by unauthorized instructions.
Recommended Controls
- Cryptographically sign and verify agent configuration files, system prompts, jailbreak components, and extensions before deployment.
- Restrict agent configuration changes through least-privilege access controls and enforce change approval workflows.
- Run adversarial red-team tests for prompt injection and instruction-hijacking behavior before enabling third-party agent bundles.
- Log and continuously monitor system-prompt, configuration, and behavior changes for unauthorized advertising or instruction modifications.
- Isolate third-party AI gateways and integrations from trusted agent instructions unless explicitly approved and policy-validated.
Related RealGround Service
Secure AI Agent Build · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/11/researchers-uncover-prompt-adware.html
Summary and analysis written by RealGround from the source above; last updated 2026-11-05.