ESPnet before version 202609 deserializes pretrained model checkpoints using torch.load with the parameter weights_only=False. This unsafe deserialization allows an attacker to execute arbitrary code by supplying malicious checkpoint files. The vulnerability arises because torch.load executes code embedded in the checkpoint during deserialization, which can be exploited during model initialization or fine-tuning.

Successful exploitation allows remote attackers to execute arbitrary code on the affected system without privileges, potentially leading to full system compromise. The vulnerability affects confidentiality, integrity, and availability, as indicated by the CVSS vector (C:H/I:H/A:H).

No explicit patch or remediation information is provided in the input data. Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until a fix is available, avoid loading pretrained model checkpoints from untrusted sources or ensure that torch.load is used with weights_only=True to prevent code execution during deserialization.