---
id: "CVE-2026-112310"
kind: "vulnerability"
aliases: ["CVE-2026-112310"]
title: "docling-core: crafted document files read arbitrary local images"
severity: "medium"
cvss: 5.3
epss: null
exploited: false
status: "open"
category: "package"
products: []
ecosystems: []
tags: ["docling", "document-processing", "package", "python"]
not_affected_if: ["no-embedded-export", "no-untrusted-documents"]
published_at: "2026-10-19T16:17:07+00:00"
first_seen_at: "2026-10-23T21:57:18+00:00"
updated_at: "2026-10-25T04:25:02+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-112310", "first_seen_at": "2026-10-23T21:57:18+00:00"}]
---

# docling-core: crafted document files read arbitrary local images

docling-core versions from 2.5.0 before 2.103.0 accept plain filesystem paths in ImageRef URIs, bypassing the allow_image_file_uri check. An attacker who supplies a crafted DoclingDocument JSON or YAML file can get local image files embedded as base64 on export or HTML rendering, which also reveals whether paths exist.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- Run `pip show docling-core` and check whether the version is below `2.103.0`.
- Search your code for `ImageRefMode.EMBEDDED` or HTML rendering of documents that come from untrusted users.

## Fix

- Upgrade docling-core to `2.103.0` or later.

## Timeline

- 2026-10-23T21:57:18+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-112310

## From the sources

docling-core 2.5.0 before 2.103.0 contains a local file disclosure vulnerability that allows attackers to read local images by supplying plain filesystem paths in ImageRef URIs, bypassing the allow_image_file_uri gate. Attackers can submit crafted DoclingDocument JSON or YAML that, when exported with ImageRefMode.EMBEDDED or rendered to HTML, embeds readable local image files as base64 and reveals path existence.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_