Vietnam has introduced a comprehensive artificial intelligence law that regulates systems similar to ChatGPT and other large language models, marking one of the most far-reaching national AI governance frameworks in Southeast Asia. The new policy, implemented through Decree 142 under the Law on Artificial Intelligence, requires companies to classify AI systems by risk level and comply with strict transparency, safety, and reporting obligations.
The regulation draws clear comparisons to the European Union’s AI Act, adopting a three-tier risk structure that categorizes systems as high, medium, or low risk depending on their potential impact on human life, security, or information integrity. High-risk applications face the most stringent requirements, including impact assessments addressing human rights and public interest concerns, while lower-risk tools are subject to lighter oversight. Authorities also require disclosure when users interact with AI chatbots and mandate labeling for deepfakes, which the law seeks to curb when they “deceive or manipulate” in ways that cause serious harm.
Vietnam’s approach goes further in granting regulatory powers over harmful synthetic media, explicitly banning deepfakes used for deception or misinformation, including in contexts linked to geopolitical tensions. Officials say the move is intended to address growing concerns about AI-generated content being used for fraud, propaganda, and misinformation, a trend increasingly observed across global digital platforms.
The law also introduces penalties for violations involving vulnerable populations, obstruction of human oversight in AI systems, and misuse of training data that breaches existing legal frameworks such as intellectual property or data protection laws. A national reporting portal will require companies to disclose incidents that may threaten cybersecurity, public order, or national security, reinforcing the government’s centralized control over AI deployment.
Foreign AI companies, including major global players, will be required to appoint local representatives to operate in Vietnam. While some firms have not yet responded to inquiries, the requirement underscores Hanoi’s intent to ensure tighter domestic accountability for foreign technology providers. At the same time, the policy includes incentives for domestic AI development, such as access to datasets and vouchers for data center usage, aimed at strengthening local startups and building Vietnamese-language AI systems.
Regional comparisons highlight diverging strategies across Southeast Asia. Malaysia is pursuing a more flexible “controlled openness” model that combines regulation with innovation sandboxes, while countries like Singapore and Thailand are adopting sector-specific or use-case-based rules. Analysts note that Vietnam’s framework, while aligned with global regulatory trends, is broader in scope and grants authorities significant discretion in enforcement, reflecting a stronger emphasis on data sovereignty and national control over AI infrastructure.
Companies have been given a transition period to comply, with full enforcement timelines extending to 2027 for sensitive sectors such as healthcare, finance, and education, signaling a phased but firm rollout of one of the region’s most ambitious AI governance regimes to date.