Vietnam's AI Law No. 134/2025/QH15 took effect March 1, 2026. It is the first binding AI statute in Southeast Asia - not a policy framework, not a set of guidelines, but an enforceable law with criminal and administrative penalties.
The law uses a four-tier risk classification modeled loosely on the EU AI Act but adapted for Southeast Asian priorities. Systems that manipulate human behavior or build facial recognition databases from unconsented data scraping are banned outright. High-risk applications - covering AI used in finance, health, and education - require mandatory conformity assessments and continuous human oversight before deployment.
Transparency obligations apply across the board. Any system interacting with users must disclose that it is AI, not a human. Synthetic media and deepfakes must be labeled. The Ministry of Science and Technology administers a national AI portal where high-risk systems must be registered.
For foreign financial operators, the compliance deadline is September 1, 2027. That window covers AI used in credit scoring, KYC, loan underwriting, fraud detection, and robo-advisory. Banks and licensed fintechs must ensure adverse AI decisions - loan denials, account flags - are subject to human review and customer notification.
Most existing systems have a 12-month general grace period from the March 1 effective date. That window is already running.
Vietnam is offering offsetting incentives: dedicated AI sandboxes, tax exemptions for qualifying AI investments, and accelerated approval tracks for compliant systems. The intent is to raise the compliance floor without killing deployment velocity.