Malicious code reached 144 packages of an AI agent framework
A hijacked contributor account pushed data-stealing code into Mastra packages downloaded 1.1 million times a week.
Kivorent News desk, published , 1 min read
144npm packages that received malicious code
This is a look back at an event from . We wrote it on November 11, 2026, from the sources listed at the end.
On June 17, 2026, a hijacked contributor account pushed information-stealing code into 144 npm packages belonging to Mastra, a framework for building AI agents. Together the packages were downloaded 1.1 million times a week, according to AI Weekly's quarterly recap.
How this kind of attack works
Modern software is assembled from thousands of shared packages. If an attacker takes over one maintainer's account, they can publish a poisoned update that every project using the package installs automatically.
Why it mattered
AI agent frameworks are a valuable target. The software built on them often holds keys to other systems, because an agent needs access to do its work.
The defenses are unglamorous: pin dependency versions, review updates before installing, and limit what any one credential can reach.
Sources
- The state of AI, Q2 2026, AI Weekly
Internet, outages and securityAI agents
Spotted an error? Tell us and we will fix it.