CVE-2026-108719 Details

A blind server-side request forgery (SSRF) vulnerability has been identified in LLMGateway versions through 1.20.0. This vulnerability allows API key holders to send requests to internal hosts by exploiting the video-generation callback_url feature. The issue arises because the callback URL is only validated for correct syntax, not for safety or destination. As a result, attackers can use loopback, private, or cloud-metadata URLs to reach internal services from the worker's network, bypassing security controls that are applied to other types of webhook deliveries.

The vulnerability can be remediated by adding a destination validation check to the video callback delivery process, ensuring that only safe, external HTTPS URLs are accepted and that internal addresses are blocked. This can be done by applying the same webhook URL validation guard that is used for other types of deliveries, such as the platform-webhook.