AI Code Tools And Aging Software Expose Enterprise Networks
Only 20% of AI-generated software recommendations are valid and free of known security flaws, according to a new industry study.
- By Jesse Jacobs
- Nov 17, 2026
Enterprise security teams face mounting supply chain risks as developers rely on artificial intelligence tools and outdated third-party components to build business software, according to a new industry report.
The State of Software Dependency Trust 2026 report, published by CleanStart, found that only 20% of tested AI-generated software dependency recommendations were both valid and free of known vulnerabilities.
Audited business applications averaged 1,180 open-source components, with 64% introduced indirectly through nested third-party tools. Nearly seven in 10 components (68%) were more than two years old, and 41% trailed current versions by 10 or more updates.
Across 947 audited commercial codebases, the average number of security vulnerabilities rose 107%, climbing from 280 to 581. Despite the surge in flaws, roughly 95% of vulnerable component downloads examined in telemetry data already had a patched version available at the time of download.
The study also highlighted governance blind spots across corporate environments:
- Shadow components: 17% of observed software components sat outside conventional tracking tools.
- Maintainer bottlenecks: A review of 47 major open-source projects found that a small group of developers handled most updates in 40% of those projects.
- AI safeguards: Connecting AI coding tools to live registry data reduced high- and critical-risk recommendations by roughly 70%.
Researchers noted that compromised publishing credentials, malicious installation scripts and manipulated build pipelines continue to threaten corporate infrastructure, making it critical for security leaders to verify software origins and integrity before deployment.