The vulnerability arises from the _load_model() function in neural_magic_training.py of the optimate project, which loads model files using torch.load() without the weights_only=True security parameter. This insecure deserialization allows an attacker to craft a malicious model file that, when loaded, executes arbitrary Python code via Pickle deserialization. The issue was identified in a commit dated 2024-07-21 and published as CVE-2026-31219. There is no CVSS score or vendor advisory specifying remediation or patch availability. The vulnerability is local to the script usage and does not involve cloud-hosted services.

Successful exploitation allows a remote attacker who can supply a malicious model file to execute arbitrary code on the victim's system during model loading. This can lead to full system compromise depending on the privileges of the user running the script. No known exploits in the wild have been reported, and the impact is limited to environments where untrusted model files are loaded using this script.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is available, users should avoid loading untrusted or unauthenticated model files with the vulnerable script. Applying the weights_only=True parameter in torch.load() when loading models is recommended to prevent arbitrary code execution. Monitor official project repositories for updates or patches addressing this vulnerability.