Deepfakes and AI Content in Vietnam: What the New Law Regulates Starting from 2026

The rapid development of generative AI is making the creation of images, videos, audio, and text faster and more accessible than ever before. With today’s widely available tools, users can generate highly realistic content, including products that simulate the face, voice, or actions of a real person. While this technology opens up numerous opportunities for media, education, marketing, and content production, it also introduces new challenges regarding fraud, misinformation, personal rights, and accountability in AI usage.

In Vietnam, this matter has entered a new phase with the Law on Artificial Intelligence No. 134/2025/QH15, passed by the National Assembly on December 10, 2025, and officially taking effect on March 1, 2026. This dedicated AI legislation emphasizes safe, responsible, and human-centric AI development while ensuring human oversight over AI systems. Is deepfake banned in Vietnam? Must AI-generated content be labeled? And what should businesses keep in mind when adopting AI?

What is Notable About the 2025 Law on Artificial Intelligence?

The Law on Artificial Intelligence No. 134/2025/QH15 comprises 8 chapters and 35 articles, governing the research, development, provision, deployment, and use of AI systems; the rights and obligations of relevant organizations and individuals; and state management responsibilities in this domain. Taking effect on March 1, 2026, the law establishes a dedicated legal framework for AI as this technology sees increasingly widespread adoption across Vietnam.

A notable aspect is that the Law does not adopt a blanket prohibition approach toward technology. Instead, it establishes a risk-based management mechanism. AI systems are categorized into high risk, medium risk, and low risk based on factors such as their impact on human rights, safety, security, user scope, and scale of influence. Alongside this classification are requirements regarding transparency, risk management, conformity assessment, and human oversight for high-risk systems.

This approach is particularly meaningful for content generated or modified by AI. Rather than treating all AI-generated outputs as potential risks, the law focuses on controlling how the technology is developed and deployed, its intended use, and the potential impact it may cause.

Are Deepfakes banned in Vietnam?

This is perhaps the most frequently asked question regarding the new AI Law. First, it is important to understand that deepfakes are not inherently illegal in all cases. Deepfake is fundamentally a technology that leverages AI to generate or alter images, videos, and audio to simulate real people or events. This technology can be utilized in various legitimate activities such as filmmaking, research, training, or content production, provided it complies with applicable requirements.

The law specifically targets and controls acts of using AI to forge or simulate real persons or events to intentionally and systematically deceive or manipulate human perception and behavior, causing serious harm to legitimate rights and interests. The Law also strictly prohibits the creation or dissemination of fraudulent content that has the potential to cause severe harm to national security, public order, and social safety.

Therefore, one should not simply assume that “the AI Law bans deepfakes.” A more accurate interpretation is that the law sets boundaries against exploiting AI technology for forgery, deception, manipulation, or causing severe consequences. This is a critical point that businesses and AI users must keep in mind when creating or distributing content that simulates real human subjects.

Must AI-Generated content be labeled?

Beyond deepfakes, another significant update in the Law on Artificial Intelligence is the requirement for transparency and identification of AI-generated content. According to official introduction materials on the Law, transparency obligations include identifying interactions with AI and labeling content generated by artificial intelligence. Concurrently, the Law strictly prohibits concealing information subject to mandatory disclosure, transparency, or explanation, as well as erasing or falsifying mandatory information, labels, or warnings in AI-related activities.

This regulation carries substantial practical implications for media and marketing operations. Businesses can now utilize AI to produce advertising images, product introduction videos, voiceovers, virtual avatars, and various other forms of content. As AI-generated products become increasingly indistinguishable from human-created content, transparency regarding the origin of content becomes an essential factor in maintaining customer trust.

However, enterprises should not interpret the regulation as requiring identical treatment for all content involving any AI tool in every circumstance. Specific application must be evaluated based on system type, content category, intended purpose, and relevant legal requirements. When establishing internal AI workflows, businesses should monitor forthcoming implementing guidelines and technical standards to ensure full compliance.

Why does the AI Law Emphasize Human oversight?

One of the foundational principles of the Law on Artificial Intelligence is being human-centric. This is reflected in the fact that the legal framework looks beyond AI’s technical capabilities to mandate human monitoring, intervention, and control. The Law strictly prohibits obstructing, disabling, or distorting mechanisms for human oversight, intervention, and control over AI systems.

In state management and public service delivery, AI cannot replace human decision-making authority, and human decision-makers remain ultimately accountable. For high-risk AI systems or those with significant social impact, the law further mandates impact assessments, risk identification, and public disclosure of assessment reports, excluding confidential information.

For businesses, this principle translates into a highly practical requirement: AI produces outputs, but humans must still review and assume responsibility for the decisions made using those outputs. Particularly for content related to customers, branding, personal data, or high-impact business decisions, human-in-the-loop review processes remain vital.

Deepfakes and the Law on Cybersecurity 2025

The Law on Artificial Intelligence is not the sole legal document to consider regarding deepfakes. The Law on Cybersecurity No. 116/2025/QH15, passed by the National Assembly on December 10, 2025, and taking effect on July 1, 2026, adds another critical regulatory layer to cyberspace activities.

This highlights that deepfakes must be viewed within a broader legal landscape. An AI-generated asset may simultaneously intersect with cybersecurity, personal data protection, image rights, intellectual property, or other regulatory provisions depending on its content and purpose. Therefore, simply checking whether a tool can generate deepfakes is not enough; it is essential to evaluate the content created, the data utilized, and how that content is deployed in practice.

BPO.MP – Partnering with businesses in AI implementation

AI delivers true value only when deployed alongside the right people, streamlined processes, and appropriate governance mechanisms. BPO.MP partners with enterprises in AI training and practical adoption, empowering teams to accurately understand the technology, select suitable use cases, and systematically integrate AI into day-to-day operations.

Has your business established an internal policy on AI utilization? Contact BPO.MP today to discuss building safe, effective AI capabilities tailored to your operational needs.