langchain-nvidia-ai-endpoints local file disclosure via VLM image inputs

langchain-nvidia-ai-endpoints versions before 1.4.2 accepted local filesystem paths as image inputs for Vision Language Model requests, allowing attackers to read files accessible to the application process. The vulnerability affects applications using VLM image inputs with ChatNVIDIA or NVIDIARerank APIs if untrusted users can control image URLs or document metadata.

Disclosed 24 September 2026 · Record updated 1 October 2026

Impact

Local file disclosure if applications accept attacker-controlled image inputs for VLM requests

Our coverage

A robot's eye ingests a folder from a locked filing cabinet, then transmits its contents through an open window to an unseen receiver outside.

Vulnerabilities

NVIDIA LangChain package leaked local files via image inputs

A GitHub advisory published on 24 September 2026 says langchain-nvidia-ai-endpoints accepted local file paths as vision model images, sending them to a remote endpoint.

24 Sept 2026

Sources

github.comhttps://github.com/advisories/GHSA-g28h-2cmm-rj9x