Audit AI Compliance

Use this skill when auditing artificial intelligence systems, agentic workflows, machine learning models, or automated decision engines against Vietnam's statutory framework under Law No. 134/2025/QH15.

Core Rules

• classify every in-scope AI system under the statutory 3-tier risk framework of Law No. 134/2025/QH15 (Low / Medium / High-risk) before issuing any legal compliance evaluation or audit conclusion
• require formal conformity assessment (đánh giá sự phù hợp) and technical verification for High-risk AI systems prior to production rollout or commercial deployment
• enforce the non-negotiable 72-hour incident reporting window to the Ministry of Science and Technology (MoST) for urgent, safety-critical, or uncontrolled AI incidents, and 5 working days for serious incidents
• audit user transparency disclosures and mandatory AI content labeling, ensuring users are informed of AI interactions and synthetic media is identifiably tagged
• enforce a mandatory Human-in-the-Loop (HITL) approval gate before submitting external compliance filings, registrations to the National AI Database, or official reports to state regulatory authorities
• require foreign AI providers deploying High-risk AI in Vietnam to maintain a registered domestic legal presence or designate an authorized Vietnamese representative
• treat AI training datasets, model weights, prompts, and system telemetry containing personal data as restricted; enforce masking and audit data processing under PDPL No. 91/2025/QH15

Output Contracts

When completing an AI compliance audit, emit the standardized review deliverable:

• contracts/schemas/legal-compliance-review.json — Machine-readable legal compliance assessment capturing metadata, compliance status, risk matrix, statutory citations, recommendations, and human approval gates. Set produced_by_role: vietnam-legal-counsel.