Key takeaways

Below is our summary of key issues under the draft AI Law:

• Implementation timeline: The draft AI Law is scheduled to enter into force on 1 January 2026, superseding AI rules in the Digital Technology Industry Law. However, the law will be rolled out in phases: the provisions on prohibited acts and regulatory sandbox will apply after 12 months from the law's effective date, while the obligations for high-risk AI systems will only take effect after 18 months.
• Foundational principles: All activities related to AI systems in Vietnam must adhere to principles such as fairness and non-discrimination, transparency and explainability, human centrism, security and safety, and accountability.
• Risk-based classification and governance: AI systems are categorized into four levels: unacceptable (prohibited), high (strict obligations like assessments and registration), medium (transparency requirements), and low (voluntary standards) risks. Businesses in sectors like finance, health, or justice face heightened scrutiny.
• General-purpose AI (GPAI) model: Modeled after EU concepts, GPAI providers must meet basic obligations (e.g., documentation, IP compliance) and enhanced ones for systemic risks (e.g., evaluations, cybersecurity). This will have impacts on tech firms integrating large models.
• Role-driven accountability: Supply chain stakeholders, including AI developers, providers, deployers, and users, may hold different responsibilities depending on their actual role and control over an AI system, which will be further elaborated by the Government under a subordinate legal instrument. Joint liability for compensation may also arise where the roles of the parties cannot be clearly delineated.
• Incentives and Supports for Innovation: Establishes a National AI Development Fund, sandboxes for startups/SMEs, AI clusters, and tax incentives – offering opportunities for domestic and foreign businesses to invest in R&D and infrastructure.
• Revenue-based penalty and other enforcement measures: In cases of serious violations committed by a large-scale organization, the monetary fine imposed could be calculated based on a percentage of the organization's global revenue from the previous fiscal year. Measures such as temporary termination or permanent prohibition of an AI system can also be applied if the system is found to cause a serious national security risk, among other instances.

Being fast tracked to meet this year's deadline, the draft AI Law will be discussed and passed by the National Assembly in its upcoming meeting session starting from 20 October 2025. Time is of the essence here – stakeholders need to closely track the drafting progress and quickly submit their comments to the lawmakers.

* * * * *

Alex Do, IPTech Executive and Danny Truong, Trainee have contributed to this legal update.