Return to Threats

Study Finds Continued Use of AI Models Flagged for Arbitrary Code Execution

arXiv 2026-10-12 AI supply chain Critical
Run AI Security Assessment Talk to AI CISO

Executive Summary

An arXiv study found a model flagged for arbitrary code execution through unsafe deserialization still referenced 594 times across 449 repositories.

What Happened

An empirical study examined vulnerable models in the AI supply chain, including a model flagged for arbitrary code execution through pickle deserialization. The model appeared in 594 references across 449 repositories, indicating continued downstream adoption despite the warning.

Vulnerability / Attack Technique

Unsafe model file formats in the AI supply chain

Technical Mechanism

Loading certain model file formats can execute embedded code; flagged models remain in use.

AI Security Category

LLM03:2025 Supply Chain

Part of the AI Supply Chain and AI SBOM topic.

Relevant Frameworks

LLM03:2025 Supply Chain

RealGround Analysis

Warnings on model hubs are not stopping reuse. RealGround analysis: prefer safe model formats and scan models before loading.

Recommended Controls

  • Use safetensors or other non-executable formats
  • Scan downloaded models before loading

Related RealGround Service

AI Agent SBOM and LLM Mapping · or start with the free AI Agent Security Scanner.

Sources

arXiv: https://arxiv.org/html/2610.12893v1

Summary and analysis written by RealGround from the source above; last updated 2026-10-18.

Talk to AI CISO