gcdRadar
archive/Research/AIR-2025-0054
closedclosed 6 May 2025Research1 Research / proof of conceptAI

AI Model Tampering Risk via Exposed Container Registries

Security researchers identified thousands of misconfigured private container registries exposing unauthenticated AI model files. This exposure left sensitive models vulnerable to unauthorized inspection and malicious modification.

first report
22 Apr 2025
last update
22 Apr 2025

Timeline · 5

  1. 22 Apr 2025AI Model Tampering via Supply Chain Attack
  2. 22 Apr 2025Silent Sabotage: Weaponizing AI Models in Exposed Containers
  3. 22 Apr 2025Exposed Container Registries: A Potential Vector for Supply-Chain Attacks
  4. 22 Apr 2025Mining Through Mountains of Information and Risk: Containers and Exposed Container Registries
  5. 22 Apr 2025The Growing Threat of Unprotected Container Registries: An Urgent Call to Action

Sources · 5

News and write-ups

  1. 2025-04-22
    Silent Sabotage: Weaponizing AI Models in Exposed Containers
    Trend Micro · trendmicro.com/vinfo/br/security/news/cyber-attacks/silent-sabotage-weaponizing-
  2. 2025-04-22
    Exposed Container Registries: A Potential Vector for Supply-Chain Attacks
    Trend Micro · trendmicro.com/vinfo/us/security/news/virtualization-and-cloud/exposed-container
  3. 2025-04-22
    Mining Through Mountains of Information and Risk: Containers and Exposed Container Registries
    Trend Micro · trendmicro.com/vinfo/us/security/news/virtualization-and-cloud/mining-through-mo
  4. 2025-04-22
    The Growing Threat of Unprotected Container Registries: An Urgent Call to Action
    dreher.in · dreher.in/blog/unprotected-container-registries

Research and catalogues

  1. 2025-04-22
    AI Model Tampering via Supply Chain Attack
    MITRE ATLAS · atlas.mitre.org/studies/AML.CS0028

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

Classification

AI role
Attack on AI
categories
Supply chainInfrastructure vulnerability
severity
1 · Research / proof of concept
targets
Open-weight model
owasp llm top 10
LLM03:2025 Supply Chain
dates
happened 26 Sep 2023 · first reported 22 Apr 2025
origin
MITRE ATLAS
Derived from MITRE ATLAS (Apache 2.0, © MITRE).
record
confidence 0.90 · created 2026-10-18 06:55Z · updated 2026-10-18 06:55Z

Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned

Copy a brief

This record as a short Markdown brief, for a ticket, a chat or a coding agent.

the brief as text
# AI Model Tampering Risk via Exposed Container Registries

https://gcdradar.vercel.app/incidents/ai-model-tampering-risk-via-exposed-container-registries · AIR-2025-0054 · closed 2025-05-06

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2025-04-22 (happened 2023-09-26)
- AI role: Attack on AI

## Summary

Security researchers identified thousands of misconfigured private container registries exposing unauthenticated AI model files. This exposure left sensitive models vulnerable to unauthorized inspection and malicious modification.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2025-04-22: The Growing Threat of Unprotected Container Registries: An Urgent Call to Action (https://dreher.in/blog/unprotected-container-registries)
- 2025-04-22: Mining Through Mountains of Information and Risk: Containers and Exposed Container Registries (https://trendmicro.com/vinfo/us/security/news/virtualization-and-cloud/mining-through-mountains-of-information-and-risk-containers-and-exposed-container-registries)
- 2025-04-22: Exposed Container Registries: A Potential Vector for Supply-Chain Attacks (https://trendmicro.com/vinfo/us/security/news/virtualization-and-cloud/exposed-container-registries-a-potential-vector-for-supply-chain-attacks)
- 2025-04-22: Silent Sabotage: Weaponizing AI Models in Exposed Containers (https://trendmicro.com/vinfo/br/security/news/cyber-attacks/silent-sabotage-weaponizing-ai-models-in-exposed-containers)
- 2025-04-22: AI Model Tampering via Supply Chain Attack (https://atlas.mitre.org/studies/AML.CS0028)

## Sources

- 2025-04-22 MITRE ATLAS: AI Model Tampering via Supply Chain Attack https://atlas.mitre.org/studies/AML.CS0028
- 2025-04-22 Trend Micro: Silent Sabotage: Weaponizing AI Models in Exposed Containers https://trendmicro.com/vinfo/br/security/news/cyber-attacks/silent-sabotage-weaponizing-ai-models-in-exposed-containers
- 2025-04-22 Trend Micro: Exposed Container Registries: A Potential Vector for Supply-Chain Attacks https://trendmicro.com/vinfo/us/security/news/virtualization-and-cloud/exposed-container-registries-a-potential-vector-for-supply-chain-attacks
- 2025-04-22 Trend Micro: Mining Through Mountains of Information and Risk: Containers and Exposed Container Registries https://trendmicro.com/vinfo/us/security/news/virtualization-and-cloud/mining-through-mountains-of-information-and-risk-containers-and-exposed-container-registries
- 2025-04-22 dreher.in: The Growing Threat of Unprotected Container Registries: An Urgent Call to Action https://dreher.in/blog/unprotected-container-registries

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.

Related

  1. closed8.8
    llama.cpp CVE-2025-49847: Buffer overflow in llama.cpp vocabulary loading

    Patch now · Disclosed · since 17 Jun 2025 · CVE-2025-49847

    AI
  2. closed7.3
    llama.cpp CVE-2026-78147: llama.cpp RPC server vulnerable to insecure deserialization

    Patch now · Disclosed · since 23 Aug 2026 · CVE-2026-78147

    AI
  3. closed7.5
    vLLM CVE-2026-41523: vLLM Security Check Bypass via assert Statement

    Patch now · Disclosed · since 16 Jun 2026 · updated 22 Jun 2026 · CVE-2026-41523

    AI
  4. closed7.6
    NVIDIA NeMo CVE-2025-23249: Deserialization Vulnerability in NVIDIA NeMo Framework

    Patch now · Disclosed · since 22 Apr 2025 · CVE-2025-23249

    AI
  5. closed9.6
    llama-cpp-python CVE-2024-34359: Llama-cpp-python remote code execution via Jinja2 chat templates

    Patch now · Disclosed · since 13 May 2024 · updated 14 May 2024 · CVE-2024-34359

    AI
  6. closed7.8
    NVIDIA NeMo CVE-2025-23303: NVIDIA NeMo Framework Deserialization Vulnerability

    Patch now · Disclosed · since 13 Aug 2025 · CVE-2025-23303

    AI