HighVulnerability

CVE-2025-9905: Keras Model.load_model arbitrary code execution through crafted .h5 archives

Source: NVD(opens in a new tab)
Identifier: CVE-2025-9905
Published: September 19, 2025
Record updated: October 9, 2026

Affected: keras >= 3.0.0, < 3.11.3
Fixed in: 3.11.3
released August 21, 2025, 29 days before disclosure
Known exploitation: Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
EPSS: 0.2%

Summary

CVE-2025-9905 is a flaw in the Keras Model.load_model method that allows arbitrary code execution even when safe_mode=True is set. A specially crafted .h5/.hdf5 model archive uses a Lambda layer to embed pickled Python code, and the safe_mode=True option is not honored when reading .h5 archives. The .h5/.hdf5 format is a legacy format that Keras 3 supports for backwards compatibility.

Mitigation

The source does not state a fix yet. Check the original advisory for updates.

Details

Scores and exploitation

CVSS score: 7.3(high)
Patch available: Yes
EPSS: 0.2%Estimated probability that this CVE is exploited in the next 30 days, from the FIRST Exploit Prediction Scoring System. Read on November 4, 2026.

Weaknesses and techniques

Weakness (CWE): CWE-913
AI threat technique (MITRE ATLAS): AML.T0010

Affected

Packages: keras@>= 3.0.0, < 3.11.3 (fixed: 3.11.3)
Vendors: Google

Classification

Attack type: Model poisoning,
Supply chain
Attack sophistication: Moderate
AI component targeted: Framework
Impact: Confidentiality,
Integrity,
Availability
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. The source gives this advisory a description and no title, so the title above is a headline generated from the source text.

Original source: https://nvd.nist.gov/vuln/detail/CVE-2025-9905

First tracked: February 15, 2026 at 08:42 PM ET, from NVD/CVE Database

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

Classifier confidence: 95% (self-assessed)

Affected packages in the Exposure Registry

Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.

kerasPyPINo LLM dependency detected · 6 tracked dependents

Summary

Mitigation

Affected packages in the Exposure Registry

Related items