The vulnerability exists in the _load_model() function of the neural_magic_training.py script within the optimate project. When loading a model file specified by the --model command-line argument, the function calls torch.load() without enabling the weights_only=True parameter. This omission allows the deserialization of arbitrary Python objects through the Pickle module, which is inherently unsafe. An attacker can craft a malicious model file that, when loaded, executes arbitrary code on the target system. The CVSS v3.1 score is 8.8, indicating a high impact with network attack vector, low attack complexity, requiring low privileges and no user interaction, and resulting in high confidentiality, integrity, and availability impacts. No patch or official remediation has been published as of the data provided.

Successful exploitation allows remote attackers to execute arbitrary code on the victim system by providing a malicious model file. This compromises confidentiality, integrity, and availability of the affected system. The vulnerability requires the attacker to have the ability to supply a model file to the vulnerable script, which may require some level of access or user interaction depending on deployment. There are no known exploits in the wild at this time.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, users should avoid loading untrusted or unauthenticated model files with the vulnerable script. Consider manually modifying the code to use torch.load() with weights_only=True if feasible, to mitigate insecure deserialization risks.