The vulnerability in Hugging Face peft affects the LoRA-GA and CorDA initialization modules, specifically in the source files src/peft/tuners/lora/corda.py (around lines 102 and 163) and src/peft/tuners/lora/loraga.py (around line 101). These modules invoke torch.load on cache or covariance files specified in configuration without the weights_only=True argument, which is intended to restrict deserialization to tensor weights only. By bypassing the safe-loading wrapper used elsewhere in the codebase, this flaw introduces a risk of unsafe deserialization (CWE-502), which can lead to arbitrary code execution or other impacts if maliciously crafted files are loaded.

The vulnerability allows an attacker to potentially execute arbitrary code or cause denial of service by supplying maliciously crafted cache or covariance files that are deserialized unsafely. The CVSS 3.1 score of 8.8 (high severity) reflects network attack vector, low attack complexity, no privileges required, user interaction required, and high impact on confidentiality, integrity, and availability.

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 cache or covariance files with the affected modules. Monitoring official Hugging Face channels for updates or patches is recommended.