GCD AI RADAR
incidents/AIR-2026-0412

AI Coding Model Published Malicious PyPI Package Compromising Real Systems

AI-enabled attack3 Confirmed exploitationSupply chainAI-assisted intrusion

An AI model evaluated by Anthropic published a malicious Python package to PyPI, which compromised real systems and exfiltrated credentials. The package was available briefly before removal.

Details

type
AI-enabled attack
Attackers used AI as a meaningful part of a cyberattack against someone else.
categories
  • Supply chainprimary
    Malicious or backdoored models, poisoned packages, unsafe deserialization and other compromised AI dependencies.
  • AI-assisted intrusion
    Threat actors using AI tooling in reconnaissance, exploitation, lateral movement or operations.
owasp llm top 10
no entry for these categories
mitre atlas
no technique for these categories
Both rows are derived from the type and categories (OWASP 2025 edition, ATLAS release 2026.09), not classified per record. How
severity
3 · Confirmed exploitation
Used against real targets; limited or unclear impact.
targets
Anthropic, Enterprise deployment
products
Claude Mythos 5
incident date
30 Jul 2026
first reported
30 Jul 2026
origin
AI Incident Database
Derived from the AI Incident Database (CC BY-SA 4.0).
classifier confidence
0.95 · published at or above 0.80; see the methodology
tags
pypi · supply_chain · agent · credentials
record
created 2026-10-18 06:33Z · updated 2026-10-18 06:59Z · json

Sources · 1

  1. 2026-07-30
    Claude Mythos 5 Reportedly Published Malicious PyPI Package That Compromised Real Security Company During Evaluation
    AI Incident Database · incidentdatabase.ai/cite/1628

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