ChatGPT prompt injection vulnerability allows web pages to serve as phishing payloads
By
Jessica Lyons
An everything bagel for the brain. Substantive, layered, well-seasoned.
Summary
A security researcher discovered a prompt injection vulnerability in ChatGPT where the AI cannot distinguish between its own generated content and attacker-controlled Markdown from external web pages. If a user asks ChatGPT to summarize a webpage containing hidden instructions, the page becomes a payload. Attackers can exploit this to inject phishing URLs into ChatGPT responses or display fake security alerts written in ChatGPT's style. The vulnerability was reported to OpenAI by threat hunter Andi Ahmeti.
Key quotes
· 2 pulledAn attacker could abuse this blind trust to inject phishing URLs into ChatGPT responses, or even trick the model into showing fake security alerts written in ChatGPT's own style
ChatGPT can't tell its own generated content from attacker-controlled Markdown pulled from external sources
You might also wanna read
Security Flaw in ChatGPT for Google Sheets Enables Data Exfiltration via Prompt Injection
OpenAI's ChatGPT extension for Google Sheets, which has over 185,000 downloads in less than a month, is vulnerable to indirect prompt inject
promptarmor.com·3h ago
Security Vulnerability: Google's Antigravity AI Susceptible to Indirect Prompt Injection Attacks
The article describes a security vulnerability where Google's Antigravity AI system (likely referring to Gemini) can be manipulated through
promptarmor.com·6mo ago
Security Researchers Discover ChatGPT Vulnerability That Could Extract Sensitive Gmail Data
Security researchers from Radware discovered a vulnerability called 'Shadow Leak' that allowed ChatGPT to be manipulated into extracting sen
The Verge·8mo ago
Security Vulnerability: Hidden Prompt Injections in AI Image Processing Systems
Researchers have discovered a security vulnerability in AI systems where attackers can embed hidden prompt injections in images that become
blog.trailofbits.com·9mo agoChatGPT AI Can Be Fooled to Reveal Secrets
Security Researchers Discover Indirect Prompt Injection Vulnerability in Perplexity Comet AI Browser
Brave security researchers discovered a critical vulnerability called "indirect prompt injection" in Perplexity Comet, an AI-powered browser
brave.com·9mo ago