Trustworthy AI Governance and Compliance

Overview

This area investigates how legal duties and governance principles can become executable and testable requirements for AI systems. Comparative legal and policy analysis, model evaluation and empirical research address privacy, safety alignment, cross-border data governance and deployment accountability, turning compliance principles into observable, auditable evidence.

Methods

Regulatory tracking and comparative law: follow AI regulation and policy in Hong Kong, mainland China and other major jurisdictions, identifying legal obligations that can inform evaluation requirements.

Compliance benchmarks and model evaluation: translate legal duties into executable tests of safety, privacy and compliance expectations, including SafeLawBench and PrivaCI-Bench.

Data governance and privacy: investigate cross-border data flows, privacy-preserving computing and contextual integrity to support trusted data use in the Greater Bay Area.

AI assurance and red teaming: develop systematic testing, oversight and incident-response methods for transparent and auditable AI assurance.

Policy translation and compliance assessment: translate research evidence into policy guidance, professional standards and assessment tools, including the Hong Kong generative AI technical and application guidelines and ContextLens.

Research Questions

How can legal duties be translated into auditable evaluation standards?

How should privacy and safety be evaluated in specific contexts?

What makes an AI assurance process transparent and auditable?

What governance evidence is needed for deployment across jurisdictions?

Resources & Outputs

SafeLawBench; PrivaCI-Bench; ContextLens for GDPR and EU AI Act compliance assessment; Hong Kong generative AI technical and application guidelines. Planned programmes include the Hong Kong AI Law and Governance Observatory, GBA Trusted Data and AI Initiative, and AI Assurance and Red-Teaming Laboratory.