AI-Suggested Dependencies Create Supply-Chain Risks Before Builds

Summary

The sponsored article warns that hallucinated or unvetted AI-suggested packages can enable supply-chain attacks. It describes a reported react-codeshift spread across repositories and recommends controlling dependency intake before builds.

Key points

The article describes slopsquatting as attackers registering nonexistent package names suggested by AI, then publishing malicious packages under those names.
It says researchers tracked the npm package react-codeshift across 47 AI-generated agent skills in one commit and more than 230 repositories through forks.
The article cites a USENIX Security study of 16 code-generation models and more than 500,000 code samples, reporting that some suggested names were nonexistent and nearly half of suggestions resolving to real packages had known CVEs or outdated releases.
Recommended controls include blocking direct access to unvetted public registries, isolating new dependencies for analysis, and using a governed package-ingestion gateway.
The article cites a CodeRabbit review finding 70% more defects in AI-co-authored contributions across 470 open-source pull requests, and Kusari report figures showing 9% of organizations had dedicated AI AppSec controls.
ActiveState claims its curated catalog reduces CVE exposure by roughly 95%; the article provides no independent testing for that vendor claim.

Article Details

Defense Focus: Prevent hallucinated or malicious open-source dependencies from entering developer environments and build pipelines by governing package selection at intake.
Detection Methods: Analyze newly introduced dependencies in an isolated sandbox for reachability and known vulnerabilities before allowing them into primary branches.
Pre-vet package names and sources against malicious typosquats and slopsquatting targets.
Data Sources: Public package registry entries and package-name resolution
Dependency vulnerability information, including known CVEs and outdated releases
Platforms: Developer workstations
CI/CD pipelines
AI agent environments
Defensive Actions: Block developer workstations and AI agents from directly querying unvetted public package registries during code completion.
Route newly introduced dependencies through an isolated analysis step before they reach primary branches.
Use a curated, pre-vetted package ingestion source between public registries and developer tools.
Apply package-ingestion controls before a dependency triggers a build.

MITRE ATT&CK

T1195.001 · Compromise Software Dependencies and Development ToolsThe article describes attackers registering hallucinated package names on PyPI or npm, uploading a malicious payload, and waiting for CI/CD builders to fetch the package.

Vendors

ActiveStateThe article argues that enterprises must govern package selection before code reaches the build stage, with ActiveState positioning its curated catalog as a way to block risky dependencies at intake.

Products

Curated CatalogThe article argues that enterprises must govern package selection before code reaches the build stage, with ActiveState positioning its curated catalog as a way to block risky dependencies at intake.Secure Open Source LibraryThis ingestion layer is precisely where ActiveState’s Secure Open Source Library and Curated Catalog operates.