NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

China issued a digital ID to virtual humans, then set them to work

Beijing began registering AI digital humans as identifiable entities, while brands found they could sell around the clock without a human on camera.

2026-11-08 · 824 words · NeuroAI
China issued a digital ID to virtual humans, then set them to work

In a Shanghai venue, a singer named Yuri performed for a packed crowd. She is about a year old, has never slept, and is not a person. Yuri is a virtual human (数字人) — and in 2025 she became one of the first in China to receive a government-issued digital identity, complete with a name, birth date, and registered face.

The episode captures a wider shift: China is both giving digital humans official standing and putting them to work.

What a virtual human is

A virtual human (数字人) is a computer-generated persona that can speak, gesture, and — increasingly — respond in real time. The technology blends face modeling, voice synthesis, lip-sync prediction, and posture capture. Early versions read the news from a script. Newer ones answer questions in a livestream, switching products mid-sentence.

They show up across broadcasting, e-commerce, short dramas, and public-service information. A 2025 report from the Communication University of China estimated that at least 1.3 million Chinese enterprises touch the digital-human field, and the market was heading toward about 480 billion RMB (roughly $67.6 billion / HK$528 billion) by 2025.

The government ID experiment

Beijing launched a pilot to register digital humans. Yuri's official identity let her team open a digital wallet and sign commercial deals in fashion, gaming, and automotive — making her, reports said, China's first licensed AI artist.

The move is not only symbolic. Supporters argue registration protects intellectual property and helps trace bad actors. Critics note the risk: an impersonated "doctor" digital human once pushed fake medical advice. Because a digital human cannot be sued, the logic goes, the company behind it must be accountable — and registration is how you find that company.

Where the money showed up

The commercial case arrived through livestream commerce. On Taobao, during the 2026 "618" shopping festival, digital-human livestreams reportedly averaged more than 100 million RMB in daily GMV (about $14 million / HK$110 million), drew over 1 billion views (up 180% year on year), and the nighttime slots alone lifted merchants' business by over 30%. The pattern: humans host the big launches; digital humans hold the late shift and catch demand that used to leak after midnight.

Baidu's digital-human platform, now called Baidu Yijing (formerly Huiboxing), was upgraded at the company's Create 2026 event in May 2026 into a full-scenario, multi-agent system spanning livestream, video, and real-time interaction, with an overseas version. Baidu cited average livestream conversion up 29%, delivery efficiency up 150%, and production cost down 63.3%, serving more than 100,000 clients across 30-plus industries.

Beyond the sales floor

Livestream commerce is the loudest use, but not the only one. Newsrooms clone anchors to file breaking updates at any hour; government service desks deploy digital humans to answer routine citizen questions; training and education use them for repeatable, patient instruction. The common thread is repetitive, always-on communication — exactly the work humans tire of and machines do without fatigue. That is also where the risk concentrates: the more a synthetic persona stands in for a trusted voice, the more its registration and labeling matter.

The trust problem

Registration solves only part of the puzzle. A digital human that looks convincing can still say anything its operator scripts — including things a real spokesperson would never approve. That is why China's separate content-labeling rule matters here too: a synthetic persona in a livestream is AI-made media and, under the September 2025 measures, must be marked. The ID pilot and the labeling rule are two halves of the same bet — that you can let machines stand in for people, provided you always know which machine, and which company, is behind them.

What readers can do now

  • Pilot a night shift. If you run livestream commerce, test a virtual human (数字人) for off-peak hours before committing to a full day.
  • Demand the ID. When buying digital-human services, ask whether the provider supports registration and traceability — it is becoming a trust signal.
  • Plan for disclosure. Pair any synthetic persona with a clear label; China's content-labeling rule already expects AI-made media to be marked.

Honest limitations

The 1.3 million-enterprise figure and 480 billion RMB market size come from a university report and industry estimates, not a government census, so treat them as orders of magnitude. Taobao's 618 digital-human GMV and Baidu's percentage gains are platform-reported and not independently audited; they reflect the best-case deployments, not an average. Yuri's "first licensed AI artist" status is drawn from overseas media citing Beijing's pilot, and the exact legal weight of such a registration is still emerging. Conversion and cost figures depend heavily on the category and the quality of the underlying model.

Key takeaways

  • China began issuing government-registered digital identities to virtual humans (数字人), with Yuri among the first licensed AI artists.
  • Livestream commerce showed the clearest payoff: round-the-clock selling and a measurable lift in nighttime revenue.
  • Platforms like Baidu's Yijing report large gains in conversion and cost, but the numbers are vendor-reported.

Related coverage

Cite this article

NeuroAI (脑机网), "China issued a digital ID to virtual humans, then set them to work," 8 November 2026, https://neuroai.site/a/na-aigc-virtual-human-broadcast

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Generative Media” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories