When Anyone Can Build Software, Every Choice Matters

Agentic AI systems and tools are making software creation possible to people beyond traditional development teams. Business and marketing professionals can now use AI tools to build solutions, but it’s the agents that are choosing the infrastructure and dependencies behind them. Those choices can be difficult to assess without software expertise. In this video, Brian Fox, Sonatype’s CTO and co-founder, explores why organizations need governance that helps guide AI toward better choices from the start.

glyph bracket orange
glyph bracket orange

Frequently Asked Questions

How is AI changing software development?

AI is opening software creation to anyone with an idea and access to an agentic AI system. These tools can help build a solution and choose the infrastructure and open source dependencies behind it, bringing technical decisions into workflows led by people who may not have a development background.

Why do AI-selected dependencies need oversight?

An AI agent may recommend a dependency or upgrade without current information about vulnerabilities, available versions, or compatibility. In Sonatype’s study of nearly 37,000 open source dependency upgrade recommendations, the best ungrounded models still fabricated about 6% of recommendations. More cautious models often advised making no change, leaving hundreds of Critical and High vulnerabilities unresolved in the study. The findings show why AI recommendations need to be checked against current software intelligence.

What are the best practices for governing agentic AI systems in software development?

Best practices for governing agentic AI systems include setting clear limits on what agents can access and change, applying security and policy checks when they select dependencies, and validating proposed changes before they merge. Monitor agent activity and dependencies over time. To put these practices into action for open source dependencies, use Sonatype Guide to bring component intelligence and policy guidance into development workflows, helping developers and AI agents make informed, policy-aligned choices.