AI Sec Watch: A Security Intelligence Platform for AI Systems

Luu, T.J.

LowVulnerability

CVE-2026-105742: Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI…

Source: NVD(opens in a new tab)
Identifier: CVE-2026-105742
Published: October 5, 2026
Record updated: October 9, 2026

Summary

Docling versions 2.95.0 through 2.132.0 have a flaw in the HTML image resource loader in docling/backend/utils/image_resource_loader.py. When enable_remote_fetch=True and fetch_images=True, the loader sends headers set in HTMLBackendOptions.headers to every remote image URL in an untrusted document, including across cross-origin redirects. A document author could therefore obtain the caller's configured credentials, such as API keys and cookies. The default configuration is not affected.

Mitigation

Fixed in 2.132.0.

Details

Scores and exploitation

CVSS score: 3.7(low)
CVSS vector: CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:N/A:N
Attack vector: network
Attack complexity: high
Privileges required: none
User interaction: none
EPSS: Not scored. FIRST has not published an EPSS score for this CVE yet.
Disclosure date: October 5, 2026

Weaknesses and techniques

Weakness (CWE): CWE-201
CWE-522

Classification

Attack type: Other
Attack sophistication: Moderate
AI component targeted: Framework
Impact: Confidentiality
Labels: security

How this record was produced

Summaries, mitigations and labels on this page are generated automatically from the source. The linked source is authoritative.

Original source: https://nvd.nist.gov/vuln/detail/CVE-2026-105742

First tracked: October 5, 2026 at 08:07 PM ET, from NVD/CVE Database

Classifier: claude-haiku-5-5, prompt v4

Classifier confidence: 85% (self-assessed)

Summary

Mitigation

Related items

Summary

Mitigation

Related items