Executive Summary
Permiso disclosed ChatGPhish, where ChatGPT's renderer trusts Markdown links and images in web summaries, enabling prompt injection that leads to phishing.
What Happened
Cybersecurity researchers have disclosed details of a vulnerability in OpenAI ChatGPT that leverages the artificial intelligence (AI) assistant's implicit trust in Markdown links and images to trigger prompt injections and open the door to phishing attacks. The technique has been codenamed ChatGPhish by Permiso Security. "The chatgpt.com response renderer trusts Markdown links and Markdown
Affected Products
- OpenAI ChatGPT
Vulnerability / Attack Technique
Prompt injection leading to phishing via rendered links
Technical Mechanism
As reported, injected content in summarized pages produced trusted-looking links in ChatGPT's response.
AI Security Category
LLM05:2025 Improper Output Handling
Relevant Frameworks
RealGround Analysis
Users trust links inside the assistant's answer. RealGround analysis: check link destinations even when an assistant provides them.
Recommended Controls
- Train users to verify links in AI answers
- Apply OpenAI's fixes
Related RealGround Service
Secure AI Agent Build · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/05/chatgphish-vulnerability-turns-chatgpt.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.