AIKIDO-2026-767363

openai is vulnerable to Exposure of Sensitive Information

Exposure of Sensitive Information Pre-CVE Published 3 days ago

Medium Risk

This Affects:

openai

7.6.0 - 7.7.0

Fixed in 7.8.0

Are you affected? Scan for Free

TL;DR

The openai client supports X.509 workload-identity authentication over mutual TLS, where the application supplies an Undici dispatcher and the SDK forwards the client certificate and request without independently verifying the transport. The approved-origin check does not reject credential-bearing query parameters and screens only a fixed set of sensitive headers, and target and CONNECT-proxy trust are not isolated, so the workload certificate and other authentication credentials can leave the strictly verified mTLS origin. This can expose the client certificate, private key, or exchanged access token to a proxy or an unverified endpoint. The fix adds an SDK-owned credential that enforces verified TLS, isolates proxy trust, restricts requests to the approved origin, and rejects conflicting header and query credentials.

Who does this affect?

You are affected if you are using a version that falls within the vulnerable range and you use X.509 workload-identity authentication with a caller-supplied dispatcher or CONNECT proxy.