AIKIDO-2026-285573
huggingface-hub is vulnerable to Insecure Deserialization
88
CVE Score
High
Insecure Deserialization
TL;DR
serialization/_torch.py loads Torch checkpoint files through pickle based deserialization without restricting the objects that get unpickled by default. A checkpoint file from a source that is not fully trusted can embed pickle opcodes that execute code as soon as the checkpoint is loaded. The loader also skips validating certain shard/index values before loading, letting a crafted checkpoint reach an unsafe combination of loading options. The fix switches the default to safe/restricted loading, validates the indexes, and rejects unsafe option combinations.
How do I fix this?
Upgrade the huggingface-hub library to the patch version.
Who does this affect?
You are affected if you are using a version that falls within the vulnerable range and you load a Torch checkpoint file that did not originate from a source you fully trust.
Package
Versions
0.27.0 - 1.31.0
Fixed in 1.32.0Background info
huggingface-hub is vulnerable to Insecure Deserialization in versions 0.27.0 - 1.31.0.
Get secure today,
quickly and for free.
Secure your code, cloud, and runtime in one central system. Connect a repo to discover what the reasoning agents find in your codebase.
No credit card required | Scan results in 32 seconds.
I consent to receiving marketing communications based on Aikido’s Privacy Policy.
SOC 2Compliant
ISO 27001CompliantISO 42001Compliant
FedRAMPAuthorized© All Intel data is openly available and commercially licensed.