The vulnerability exists in langchain-text-splitters before version 1.1.2, where the method split_text_from_url() validates the initial URL using validate_safe_url() but performs the HTTP fetch with redirects enabled and without revalidating redirect targets. This allows an attacker to supply a URL that redirects to internal, localhost, or cloud metadata endpoints, bypassing SSRF protections. The fetched response is parsed into Document objects and returned to the application. If the application exposes these Document contents back to the requester, sensitive internal data could be leaked. Applications that do not expose Document contents to the requester are not directly impacted by data exfiltration through this vulnerability. The issue is addressed in langchain-text-splitters version 1.1.2.

An attacker can exploit this SSRF vulnerability to cause the application to make HTTP requests to internal or cloud metadata endpoints that are normally inaccessible. If the application returns the fetched Document contents to the user who supplied the URL, this can lead to sensitive data disclosure. There is no impact on integrity or availability reported. The severity is rated medium with a CVSS score of 6.5.

Upgrade langchain-text-splitters to version 1.1.2 or later, where this SSRF vulnerability is fixed by properly validating redirect targets. Until upgraded, applications should avoid returning raw Document contents fetched from user-supplied URLs to the requester. Patch status is confirmed fixed in version 1.1.2.