AI Regulation Tracker  /  Legislation in force

Vietnam's AI Law makes any business running a chatbot tell users they are talking to a machine

Under Vietnam's new Law on Artificial Intelligence, the company that operates a customer-facing bot, not just the model maker, must ensure users can tell they are interacting with AI. The duty is already in force, with a sector grace clock running.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

A business running a chatbot in Vietnam now owes its users a sentence it may never have written before: a clear notice that they are talking to a machine. That obligation comes from the country's Law on Artificial Intelligence, No. 134/2025/QH15, which took effect on March 1, 2026. What makes the rule land on ordinary companies, rather than only on the labs that build large models, is where it places the duty. It sits on the deployer, the organization that puts the bot in front of the public.

The medium-risk trigger

As reported by law firms tracking the statute, the AI Law sorts systems into risk tiers, and human-facing interaction falls into a "medium-risk" band organized around a single idea: non-recognisability. The concern is systems that could mislead, influence, or manipulate a person precisely because the AI nature of the interaction is not readily recognisable. A support bot that answers in fluent, confident sentences, a voicebot that sounds like a person, or an AI agent that books appointments all fit the pattern. The response the law asks for is modest but firm. Users must be able to recognize that they are interacting with AI, and they must be warned when output could be confused with real events or real persons.

Note the two distinct duties bundled here. The first is baseline disclosure: tell people it is a bot. The second is narrower and sharper: when the system produces content that could pass for a genuine person or a real event, flag it. A marketing tool that generates a synthetic spokesperson, or an agent that drafts messages in a named executive's voice, carries the heavier warning obligation.

Why the "deployer" framing matters

Much of the early coverage of Vietnam's AI Law focused on providers, meaning the companies that build and supply AI systems. The transparency duty for human-facing systems is different. By attaching it to the deployer, the law reaches the retailer with an off-the-shelf support widget, the clinic with an appointment voicebot, and the brokerage with a sales agent. None of them wrote the underlying model. All of them are on the hook for telling their users what they are talking to. For a US company operating in Vietnam through a local site or app, that means the compliance task is not "did our vendor label the model" but "does our live customer experience disclose the bot."

What it does not do

The rule is a transparency measure, not a ban. It does not prohibit chatbots, voicebots, or AI agents, and it does not require a business to abandon automation. It also does not, on the reporting available, prescribe an exact wording or a single approved disclosure format, so companies retain latitude in how they phrase the notice as long as a reasonable user can tell they are dealing with AI. And it does not stand alone. It overlaps with Vietnam's earlier internet decree, Decree 147, which already requires labeling of AI-generated images, video, and audio on platforms. A business producing synthetic media may face both the AI Law's confusion warning and the decree's labeling rule at once.

The grace clock, and the wider signal

The law is in force now, but the compliance deadline is staggered. Existing deployers have until March 1, 2027 to conform, and companies in health, education, and finance have until September 1, 2027. That runway is time to fix disclosures, not a reason to defer the work, because the standard itself is already set.

For a US reader, the cross-border angle is the more telling story. Vietnam's approach mirrors the European Union's AI Act, whose transparency provisions require that people be told when they interact with an AI system, and it echoes California's bot-disclosure law, which requires certain automated accounts to identify themselves. A common norm is forming across jurisdictions: bots that talk to the public should say so. A company that builds a single, clear AI-disclosure standard into its customer channels is likely to satisfy Vietnam, the EU, and California at the same time, rather than retrofitting each market separately.

One practical caution is worth stating plainly. The specific article numbers behind Vietnam's transparency duty are not confirmed against the official gazette text, and the widely cited English translation is a paid secondary portal. The medium-risk disclosure obligation is well attested across practitioner reporting, so treat it as the operative standard, but do not build a compliance memo around a precise clause citation until the authoritative Vietnamese text is confirmed. The duty is real; the paragraph number is not yet nailed down.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.