CVE-2026-71281: peft Unsafe Deserialization via torch.load() Without weights_only in LoRA-GA and CorDA Modules
Hugging Face peft's LoRA-GA and CorDA initialization modules (src/peft/tuners/lora/corda.py lines ~102 and ~163, and src/peft/tuners/lora/loraga.py line ~101) call torch.load() on config-specified cache/covariance files without weightsonly=True, bypassing peft's own safe-loading wrapper used elsewhere in the codebase. Because torch.load() without weightsonly=True performs full pickle deserialization, loading a malicious cache or covariance file (e.g. a shared/downloaded LoRA-GA or CorDA cache) results in arbitrary code execution.
Affected Software
Event History
Frequently Asked Questions
What is the severity of CVE-2026-71281?
CVE-2026-71281 has a severity rating of 8.8, indicating a high risk level.
How do I fix CVE-2026-71281?
To fix CVE-2026-71281, ensure that the torch.load() function in LoRA-GA and CorDA modules is called with weights_only=True.
What software does CVE-2026-71281 affect?
CVE-2026-71281 affects the Hugging Face peft library.
What is the main issue with CVE-2026-71281?
The main issue with CVE-2026-71281 is unsafe deserialization due to the use of torch.load() without proper safety precautions.
When was CVE-2026-71281 published?
CVE-2026-71281 was published on August 5, 2026.