CVE-2025-1550

published 11 Mar 2025 · updated 24 Jul 2025

Arbitrary Code Execution via Crafted Keras Config for Model Loading

8.3HighCVSS 7.3EPSS 3%1 public exploits

A public exploit exists; no exploitation in the wild has been reported. EPSS puts the chance of exploitation in the next 30 days at 3%. Fixed in Keras 3.8.0.

The Keras Model.load_model function permits arbitrary code execution, even with safe_mode=True, through a manually constructed, malicious .keras archive. By altering the config.json file within the archive, an attacker can specify arbitrary Python modules and functions, along with their arguments, to be loaded and executed during model loading.

Affected

Vendors, products and packages with version ranges

VendorProductVersionsFixed
GoogleKeras>=3.0.0 <3.8.03.8.0

Exploit evidence

1 public sources

Timeline

Every event on this vulnerability

Related

Other vulnerabilities in Google, strongest first

Embed

A live severity badge for an advisory, README or ticket

CVE-2025-1550 severity on Zeptric[![CVE-2025-1550](https://zeptric.io/badge/cve/CVE-2025-1550.svg)](https://zeptric.io/cve/CVE-2025-1550)