China's state TV turns on the AI 'doctors' selling contact lenses
China spent this month putting AI faces on television; on Sunday night the state broadcaster spent an hour explaining what happens when those faces start selling medicine. The answer from the market regulator is a 5,000 yuan fine.
An AI digital human with no eyes told viewers how comfortable its contact lenses felt — and China's state broadcaster used the case to argue that the real person behind the avatar carries the liability. CCTV's weekly quality report, carried by Sina Finance and the Hunan-based Sanxiang Metropolis Daily on September 27, opens on an ad placed by an AI short-drama studio in which a virtual character says in the first person: "I wore these lenses a full day, very comfortable," with a product link attached. The line attracted immediate objection — a figure with no physiology cannot have worn anything against a cornea — and by August 6 the studio had pulled the video, days after "AI drama lead's 60-second ad slots at 258,000 yuan" trended on Weibo. The detail that lifts this out of internet-trade gossip is the product category: decorative coloured contact lenses are regulated in China as Class III medical devices, the same risk tier as implantable pacemakers, artificial joints and vascular stents. Article 16 of the Advertising Law bars medical, pharmaceutical and device advertising from using any endorser to recommend or vouch for a product, and Article 28 covers fabricated experience.
That makes the legal question sharper than "is the ad misleading". Zhu Zhenfeng of the Chinese Academy of Social Sciences' Institute of Law told CCTV that a wholly fabricated "native" AI digital human cannot be an endorser at all, because it cannot satisfy the law's requirement of genuine, sufficient use of the product. Li Huaisheng, who heads the internet-law institute at China University of Political Science and Law, framed the responsibility chain: the advertiser owns the truth of the content, publishers, endorsement agencies and the studios that build the avatars own content review, and an AI digital human "cannot itself become a shield against legal penalty" — enforcement has to pierce through to the actual parties. He also noted the labelling escape hatch does not work: a mark saying the content is AI-generated removes the liability for not labelling, not the liability for a false claim or an unlicensed product.
The second half of the report is a demonstration of how cheap the abuse is. A reporter installed one of the many apps advertising one-tap generation of digital-human sales videos — one costing 48 yuan a week — uploaded a photo of a pear, and had an eight-second clip in minutes. Prompted for "a medical worker selling a plaster", the software produced a white-coated digital human pitching plasters, with no credential check anywhere in the pipeline, and some apps in this category ship a watermark-removal function, which is a labelling-evasion feature sold as a convenience. Pan Jiming of the China Computer Federation's security committee put the framing bluntly: the technology is neutral, the problem is the people using it, and the control has to land on them. Li Xu of the University of Science and Technology of China argued the platforms should verify the identity and qualifications behind any health-adjacent digital human, and that absolute claims — "cured", "rooted out", "stop your medication", "guaranteed" — should trigger a high-risk review on their own.
The enforcement already exists, and its size is the story. On September 7 the State Administration for Market Regulation published a batch of typical online-food false-advertising cases, including the first Beijing case of an AI virtual human fined for false claims in a livestream: Beijing Shuangtaiyuan Trading Company used an AI human to tell viewers that an ordinary elderberry powder "helps boost immunity" and that "your resistance will improve after drinking it". The product was not a registered health food; the company was ordered to stop and fined 5,000 yuan, on a penalty decision dated June 8. The regulator's own framing is that an AI virtual human "is not a shield for false advertising". That case landed after February's live-commerce supervision rules took effect, and it arrived alongside a six-month nationwide crackdown on fake health claims launched in April. Treat the fine as a warning shot, not a deterrent: 5,000 yuan is roughly what a manufacturer spends to generate a batch of avatar videos.
This is the same regulatory push we covered earlier this month — China orders labels and a human sign-off on every AI broadcast — arriving from the advertising side, and it is aimed straight at the AI-drama economy we wrote about when China's prime-time TV slot belonged to an all-AI drama. The two stories are the same industry: generated personalities cheap enough to sell things, and a liability chain that nobody has yet been forced to pay for at a price that changes behaviour.
What to watch: whether the platforms add credential checks for medical personas before a bigger fine sets the real number, and whether any case reaches the studios that build the avatars rather than the retailers who rent them.
If a synthetic face cannot legally endorse anything, what is the first product an AI presenter should be allowed to sell? Tell us in the comments.
Sources: CCTV News via Sina Finance · SAMR — typical cases of online food-sales false advertising (Xinhua) · Southern Metropolis Daily via N Video — Beijing's first AI virtual-human fine · Procuratorate Daily via Sina — drawing a legal boundary for AI actors in ads
