Researcher Discovers Critical SSRF Vulnerability in ChatGPT’s Custom GPT Feature

Security researcher SirLeeroyJenkins uncovered a significant Server-Side Request Forgery (SSRF) vulnerability in ChatGPT's Custom GPT feature while testing its API integration capabilities. The flaw allows malicious actors to manipulate the system into fetching data from arbitrary URLs, potentially exposing internal networks or cloud metadata services.

The vulnerability stems from insufficient URL validation in the Actions feature, which enables Custom GPTs to interact with external APIs. "Once I realized this feature could return data from any user-provided URL, the hacker instinct kicked in," Jenkins remarked. SSRF vulnerabilities ranked among the OWASP Top 10 web security risks in 2021 and have grown more dangerous with the proliferation of insecure cloud configurations.

Jenkins' Medium post details two primary SSRF attack vectors: full-read and blind. The discovery highlights ongoing security challenges in AI platforms as they expand integration capabilities. OpenAI has not yet commented on potential mitigation timelines.