Microsoft has launched ASSERT, an open-source framework designed to simplify the process of evaluating and ensuring specific behaviors in application-specific artificial intelligence models.

The tool takes text descriptions of goals, policies or intended behaviors and turns them into comprehensive, scored tests. It then runs these against the target system, providing detailed feedback on performance.

Developers can input context, tools, and constraints to further tailor evaluations, ensuring that AI systems behave in line with business needs. For instance, a document research agent could be prevented from sending confidential emails outside the company.

The ASSERT framework addresses the gap between broader evaluations and application-specific requirements, offering continuous monitoring capabilities during system construction or after deployment. This move comes as the AI industry shifts towards more repeatable testing and regression checks.