CVE-2026-73325 Fujitsu Research CVE debrief

Who should care

Organizations using Fujitsu Research's OneCompression library before 1.2.1 should prioritize reviewing their inventory and considering upgrades to version 1.2.1 or later. Additionally, developers and security teams working with machine learning models and PyTorch should be aware of the potential risks associated with deserialization of untrusted model files.

Technical summary

The CVE-2026-73325 vulnerability is caused by the unsafe deserialization of model.pt checkpoint files in Fujitsu Research's OneCompression library before 1.2.1. The QuantizedModelLoader.load_quantized_model_pt() function calls torch.load with weights_only=False, which invokes Python's pickle machinery during deserialization. This allows attackers to execute arbitrary code by embedding malicious __reduce__ methods in a crafted model checkpoint.

Defensive priority

High-priority defensive actions are required due to the HIGH CVSS score of 8.4. Organizations should immediately review their inventory for the affected Fujitsu Research's OneCompression library before 1.2.1 and consider upgrading to version 1.2.1 or later.

Recommended defensive actions

• Review inventory for the affected Fujitsu Research's OneCompression library before 1.2.1
• Consider upgrading to version 1.2.1 or later
• Implement compensating controls such as restricting access to model directories
• Monitor for suspicious activity related to model loading and deserialization
• Exception tracking for model loading failures

Evidence notes

The CVE-2026-73325 record indicates an unsafe deserialization vulnerability in Fujitsu Research's OneCompression library before 1.2.1. The vulnerability allows attackers to execute arbitrary code by supplying a crafted model.pt checkpoint file. The QuantizedModelLoader.load_quantized_model_pt() function unconditionally calls torch.load with weights_only=False, invoking Python's pickle machinery during deserialization. Limited evidence is available about the scope of affected systems and potential mitigations.

Sources and references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.