How to Close the AI Governance Gap in Software Development

Widespread adoption of AI coding tools accelerates software development but also opens up new vulnerabilities that demand stronger governance and oversight. To close the AI governance gap, the article recommends three core components: Observability, Benchmarking, and Education, to be embedded across the repository workflow so security is built in by design.

Observability provides granular insights into code health, suspicious patterns and where AI-generated code enters the system, enabling early detection of flaws before they become attack vectors. Benchmarking encourages governance leaders to assess developers’ security aptitude and establish trust scores for learning programs, while Education raises awareness and improves code review and testing through hands-on, real-world exercises.

According to the Stack Overflow Developer Survey, three-quarters of developers are using or planning to use AI coding tools, but only 42 percent trust the accuracy of AI output in their workflows, and BaxBench reports that 62 percent of solutions from even the best model are either incorrect or contain a vulnerability. CISOs must therefore collaborate with other leaders to implement a comprehensive, automated governance plan that enforces policies and guardrails from the outset. 5 September 2025.