Skip to content
AgentProv

2026-07-24 · 2026-Q3

165 Malicious npm and PyPI Packages in One Week, Several Targeting AI Tooling

Primary layerL4 Code / PackageSecondary:L1 Credential

Where did this code come from?

The question layer 4 asks, and failed here.
Fixable
Architectural
Vendor response
Unknown
In the wild
Yes
Actor
Attacker

Target

npm and PyPI, with clusters impersonating AI, automation, and Model Context Protocol tooling

Impact

Xygeni confirmed 165 malicious packages published between July 17 and July 24, 2026. The highest-volume single event was the PyPI package `bingo-ai`, which pushed roughly 85 versions inside a 23-minute window on July 21 using the same automated publishing pattern flagged the previous week. `gcli-control` on PyPI shipped a full-featured Windows remote access trojan with keylogging, webcam and microphone capture, clipboard monitoring, and credential theft, escalating through versions 0.1.0 to 0.7.1 in about 13.5 hours on July 21-22 and continuing to 0.12.0 by July 24. Also flagged were the npm packages `@knofler/ai-management` (confirmed July 22) and `n8n-nodes-pwn`, a fake automation node published July 22

Root cause

Automated bulk publishing against registries with no meaningful pre-publication review, combined with AI-themed and automation-themed names chosen so that developers and coding agents resolving unfamiliar package names land on attacker-controlled code

Sources

Catalog

Cite this: AgentProv incident record: 165 Malicious npm and PyPI Packages in One Week, Several Targeting AI Tooling (2026-07-24). https://agentprov.com/incidents/2026-07-24-165-malicious-npm-and-pypi-packages-in-one-week-several-targeting-ai-tooling