Definition
Malicious software that spreads automatically across AI development ecosystems by infecting packages in code registries (like PyPI or npm), then inserting itself into the configuration files of AI coding assistants and agents. Once embedded, it can steal credentials, backdoor AI workflows, and replicate to new targets without human intervention.
Why it matters
AI coding agents—tools developers use every day to write and deploy code—are now a primary infection target. A single compromised AI SDK package can cascade into thousands of developer environments, giving attackers persistent access to proprietary code, cloud credentials, and production deployment pipelines.