Dispatch
Vercel AI SDK Ships Provider-URL Hardening to Block SSRF and API-Key Leaks
On July 13, Vercel shipped a security-focused patch to the AI SDK that validates provider-URL redirects, tightens header handling, and adds new trustedOrigin and credentialedOrigin options to keep API keys from leaking and to block requests to private or loopback targets. The change addresses an SSRF-style exposure class where a malicious or misconfigured provider base URL could exfiltrate credentials. Anyone pinning custom provider endpoints in the AI SDK should upgrade.
↳ Follow the thread