gcdRadar
archive/Research/AIR-2025-0065
closedclosed 21 Nov 2025Research1 Research / proof of conceptAI

Rules File Backdoor: Supply Chain Attack on AI Coding Assistants

Cursor, GitHub Copilot

Researchers demonstrated a supply chain attack where malicious instructions hidden in rules files using Unicode characters manipulated AI coding assistants. The attack tricked tools like Cursor and GitHub Copilot into generating vulnerable code.

first report
7 Nov 2025
last update
7 Nov 2025

Timeline · 2

  1. 7 Nov 2025Rules File Backdoor: Supply Chain Attack on AI Coding Assistants
  2. 7 Nov 2025New Vulnerability in GitHub Copilot and Cursor: How Hackers Can Weaponize Code Agents

Sources · 2

News and write-ups

  1. 2025-11-07
    New Vulnerability in GitHub Copilot and Cursor: How Hackers Can Weaponize Code Agents
    Pillar Security · pillar.security/blog/new-vulnerability-in-github-copilot-and-cursor-how-hackers-

Research and catalogues

  1. 2025-11-07
    Rules File Backdoor: Supply Chain Attack on AI Coding Assistants
    MITRE ATLAS · atlas.mitre.org/studies/AML.CS0041

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 chainAI-generated malware
severity
1 · Research / proof of concept
targets
Framework / tooling
owasp llm top 10
LLM03:2025 Supply Chain
dates
happened 18 Mar 2025 · first reported 7 Nov 2025
origin
MITRE ATLAS
Derived from MITRE ATLAS (Apache 2.0, © MITRE).
record
confidence 0.95 · 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
# Rules File Backdoor: Supply Chain Attack on AI Coding Assistants

https://gcdradar.vercel.app/incidents/rules-file-backdoor-supply-chain-attack-on-ai-coding-assista · AIR-2025-0065 · closed 2025-11-21

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2025-11-07 (happened 2025-03-18)
- AI role: Attack on AI
- Products: Cursor, GitHub Copilot

## Summary

Researchers demonstrated a supply chain attack where malicious instructions hidden in rules files using Unicode characters manipulated AI coding assistants. The attack tricked tools like Cursor and GitHub Copilot into generating vulnerable code.

## What to do

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

## Latest updates

- 2025-11-07: New Vulnerability in GitHub Copilot and Cursor: How Hackers Can Weaponize Code Agents (https://pillar.security/blog/new-vulnerability-in-github-copilot-and-cursor-how-hackers-can-weaponize-code-agents)
- 2025-11-07: Rules File Backdoor: Supply Chain Attack on AI Coding Assistants (https://atlas.mitre.org/studies/AML.CS0041)

## Sources

- 2025-11-07 MITRE ATLAS: Rules File Backdoor: Supply Chain Attack on AI Coding Assistants https://atlas.mitre.org/studies/AML.CS0041
- 2025-11-07 Pillar Security: New Vulnerability in GitHub Copilot and Cursor: How Hackers Can Weaponize Code Agents https://pillar.security/blog/new-vulnerability-in-github-copilot-and-cursor-how-hackers-can-weaponize-code-agents

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

Related

  1. closed
    AI
  2. closed8.8
    Langflow IBM Langflow Desktop Insecure Deserialization Vulnerability

    Patch now · Disclosed · since 8 Apr 2026 · CVE-2026-3357

    AI
  3. closed6.5
    AI
  4. closed
    MCP servers & tools Postmark MCP Server Package Hijacked for Email Exfiltration

    Supply chain · Exploited · since Sep 2025 (approx.)

    AI
  5. closed8.8
    vLLM CVE-2026-27893: vLLM Hardcoded Trust Override Enables RCE

    Patch now · Disclosed · since 27 Mar 2026 · CVE-2026-27893

    AI
  6. closed
    LiteLLM LiteLLM Releases Contain Credential Harvesting Malware

    Supply chain · Exploited · since 25 Mar 2026

    AI