Formal Methods Meet LLMs: Runtime Monitoring and Auditing for AI Compliance Throughout the Lifecycle
arXiv·medium signal
Combines formal methods with ML to enable offline auditing and online runtime monitoring of AI-enabled products throughout the development lifecycle. The framework addresses a governance gap: how to monitor AI behavior against temporal, product-specific compliance constraints both pre- and post-deployment. Proposes techniques for both first-party developers and third-party evaluators, making compliance provable rather than aspirational.