GCD AI RADAR
incidents/AIR-2023-0004

PoisonGPT Open Source Model Supply Chain Demonstration

Attack on AI1 Research / proof of conceptSupply chainData poisoning

Researchers demonstrated that an open-source language model could be poisoned and uploaded to a major model hub. The proof-of-concept highlighted supply chain risks within public model repositories.

Details

type
Attack on AI
An AI model, agent or AI-powered product was the thing exploited, abused or leaked.
categories
  • Supply chainprimary
    Malicious or backdoored models, poisoned packages, unsafe deserialization and other compromised AI dependencies.
  • Data poisoning
    Training or retrieval data manipulated to change model behaviour.
owasp llm top 10
mitre atlas
Both rows are derived from the type and categories (OWASP 2025 edition, ATLAS release 2026.09), not classified per record. How
severity
1 · Research / proof of concept
Demonstrated by researchers; no known real-world exploitation.
targets
Open-weight model, Framework / tooling
products
HuggingFace, LLM
incident date
Jul 2023 (approx.)
first reported
30 Oct 2023
origin
MITRE ATLAS
Derived from MITRE ATLAS (Apache 2.0, © MITRE).
classifier confidence
0.95 · published at or above 0.80; see the methodology
tags
poisongpt · supply-chain · huggingface
record
created 2026-10-18 06:55Z · updated 2026-10-18 06:55Z · json

Sources · 2

  1. 2023-10-30
    PoisonGPT
    MITRE ATLAS · atlas.mitre.org/studies/AML.CS0019
  2. 2023-10-30
    PoisonGPT: How we hid a lobotomized LLM on Hugging Face to spread fake news
    blog.mithrilsecurity.io · blog.mithrilsecurity.io/poisongpt-how-we-hid-a-lobotomized-llm-on-hugging-face-t

Titles and links belong to their publishers. The summary above is written by GCD AI RADAR in its own words; data is CC BY-SA 4.0.

Related incidents