ChatGPT Vulnerability Exposed Underlying Cloud Infrastructure

A SecurityWeek report details a recently patched ChatGPT vulnerability that could have exposed the AI chatbot’s underlying cloud infrastructure through an SSRF flaw in custom GPTs. The researcher, Jacob Krut, a bug bounty hunter and security engineer at Open Security, demonstrated how the flaw in the Actions section allowed specially crafted URLs to query a local endpoint tied to the Azure Instance Metadata Service and obtain an access token for the ChatGPT Azure IMDS identity.

The vulnerability was reported to OpenAI via its bug bounty program on the BugCrowd platform, and the vendor assigned it a high severity rating and patched it quickly, according to the article. It is unclear whether a bug bounty payout has been made.

Security expert Christopher Jess emphasised the risk, describing the SSRF in ChatGPT’s Custom GPT Actions as a textbook example of how small validation gaps at the framework layer can cascade into cloud‑level exposure, potentially enabling access to internal services and privileged cloud identities.