Scworld

Python Libraries in AI/ML Models Vulnerable to Metadata Poisoning

First seen 14 Jan 2026, 02:09 UTC • Theregister

•91% similarity •22.3

Ask AI about this cluster

Article Content

Include sub-articles 1 / 1

Browse articles

ThreatCluster

Vulnerabilities have been identified in popular AI and ML Python libraries used in Hugging Face models, allowing remote attackers to embed malicious code in metadata. This code executes automatically when a file with the poisoned metadata is loaded. The affected libraries include NeMo, Uni2TS, and FlexTok, developed by Nvidia, Salesforce, and Apple in collaboration with EPFL VILAB.

ThreatCluster AI

Attack Flow

11 entities, 16 inferred relationships (STIX 2.1).

Organization (1)

Palo Alto Networks

Vulnerability (2)

CVE-2025-23304
CVE-2026-22584

Malware (1)

Hydra

Affected Platform (5)

FlexTok
NeMo
PyTorch
Safetensors
Uni2TS

Tool (2)

Hugging Face
Python

Relationships

Hydra exploits CVE-2025-23304
Hydra exploits CVE-2026-22584
Hugging Face exploits CVE-2025-23304
Hugging Face exploits CVE-2026-22584
Python exploits CVE-2025-23304
Python exploits CVE-2026-22584
CVE-2025-23304 targets FlexTok
CVE-2025-23304 targets NeMo
CVE-2025-23304 targets PyTorch
CVE-2025-23304 targets Safetensors
CVE-2026-22584 targets FlexTok
CVE-2026-22584 targets NeMo
CVE-2026-22584 targets PyTorch
CVE-2026-22584 targets Safetensors
Hydra targets Palo Alto Networks
Hydra related to Uni2TS

Community

Browse all →

Tracked Entities in This Story

Hydra Palo Alto Networks FlexTok NeMo PyTorch Safetensors Uni2TS Hugging Face Python

Python Libraries in AI/ML Models Vulnerable to Metadata Poisoning

Ask AI about this cluster

Article Content

More articles in this cluster

Attack Flow

Defensive Countermeasures

Extracted Entities

Community

Continue Reading

Tracked Entities in This Story