Vietnam AI Law Article 11: labelling real people, real events
Under Vietnam's AI Law, deployers must label AI media that imitates a real person's face or voice or recreates a real event. Art and film get a softer rule.
In the English translation of Vietnam's Law No. 134/2025/QH15, deployers must attach an easily recognisable label to AI-generated audio, image or video that simulates the appearance or voice of a real person, or recreates a real event. They must also give a clear notice when content could confuse viewers about the authenticity of events or characters. Cinematographic and artistic works are labelled in a way that does not obstruct the display.
The deployer duties
The law takes effect 1 March 2026 and the Government will issue detailed rules. The table paraphrases the translation and is not legal advice.
| Content | Duty | Form |
|---|---|---|
| Could confuse authenticity of events or characters | Clear notice | Notice to viewers |
| Simulates a real person's appearance or voice | Easily recognisable label | Visible or audible label |
| Recreates a real event | Easily recognisable label | Visible or audible label |
| Cinematographic or artistic work | Label without obstructing the display | Unobtrusive label |
Where this meets a Sume workflow
Sume has an avatar product that produces talking-head video. If a customer makes one that resembles a real person, the label duty above would fall on the deployer, which is the customer. The Sume docs I read do not describe an automatic visible label, so the customer would add one in their own editing.
Placing a visible label
The captions route can burn authored cue text (text with start and end times) onto a clip without speech recognition, which is one way a caller could place a plain-text AI label. Check the Video captions page for the exact fields before building this.
Open points
- Detailed Government rules were not yet read.
- Which party is the deployer in a given case needs a lawyer.
- The translation is not the official text.
Sources
Related posts
More in Use cases
- Virtual staging 8 rooms, 3 takes each: Seedream vs Flux cost
A listing with 8 rooms and 3 staged takes each is 24 images: $1.20 on Seedream 4.5, $1.50 on Flux 2 Flex, $2.40 on Nano Banana 2 at 1K on Sume.
- Voicemail drop: pick immediate or delayed detection
ElevenLabs added voicemail detection for Twilio calls with immediate and delayed modes. Record the message ahead of time with Sume's tts_create and drop it.
- Walmart items with under 15 reviews: add a how-to-use clip
Walmart widened its Recognized Reviewer Program to items with under 15 reviews. While reviews build, a short how-to clip made with Sume can help.
- Walmart Listing Quality dashboard: fill thin image sets
Walmart's GenAI Listing Quality dashboard recommends catalog fixes. When the fix is more images, here is how to generate and check them with Sume before upload.
Written by Sume