CVE-2026-31222 describes an insecure deserialization vulnerability (CWE-502) in the snorkel library's Trainer.load() method. The method loads model checkpoint files using torch.load() without enabling the weights_only=True parameter, which restricts deserialization to tensor weights only. This omission allows deserialization of arbitrary Python objects through Pickle, enabling remote attackers to execute arbitrary code by providing crafted model files. The vulnerability affects versions through 0.10.0 and has a high CVSS score of 8.8. No patch or official remediation level has been published as of the data provided.

Successful exploitation allows remote attackers to execute arbitrary code on the victim system by loading a maliciously crafted model checkpoint file via the vulnerable Trainer.load() method. This can lead to full compromise of the affected system with high confidentiality, integrity, and availability impact.

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 model checkpoint files with the Trainer.load() method. Consider manually modifying the code to use torch.load() with weights_only=True if feasible to restrict deserialization. Monitor vendor channels for updates or patches.