China's AI microdrama industry has spawned an unusual labor market: platforms that pay ordinary people to license their own faces. Shenzhen-based ActID pays 99–500 yuan (roughly $15–$74) per episode for facial likenesses used in AI-generated short dramas and ads; Chengdu-based New Claw pays a minimum of 500 yuan per licensed image. Producers browse catalogs of faces sorted by age, gender, and archetype — "girl-next-door," "rugged," "supermodel" — and license whichever face fits a thriller, romance, or ad spot (Rest of World). Over 95% of microdramas released in China in early 2026 used AI in production, according to the same reporting, meaning this face-licensing marketplace is no longer a niche experiment — it is becoming standard infrastructure for a mainstream content industry.
The consent framework already exists
The reflexive read is that China's face-rental economy is running ahead of the law. It isn't. The Personal Information Protection Law (PIPL), in force since November 2021, classifies biometric characteristics as "sensitive personal information" under Article 28, processable only with "a specific purpose and a need to fulfill" plus strict protective measures. Article 29 goes further, requiring the individual's separate consent — a higher bar than PIPL's baseline consent standard (PIPL translation, Article 28–29). Layered on top, the Cyberspace Administration of China's Provisions on the Administration of Deep Synthesis of Internet Information Services — issued jointly with the Ministry of Industry and Information Technology and the Ministry of Public Security, effective January 10, 2023 — specifically require providers offering facial or voice-editing functions to notify users and obtain separate consent from the person whose biometric data is being altered (CAC official announcement).
So when ActID or New Claw structure their business as an explicit, compensated licensing transaction — a person opts in, uploads or sits for a scan, and gets paid per use — that is arguably PIPL's consent architecture functioning as designed, not evading it. A voluntary, priced market for likeness rights is a defensible outcome for a law built around individual control over sensitive data, and it beats the alternative regulators in other jurisdictions have reached for: blanket bans on commercial biometric processing that leave no lawful channel for people who want to monetize their image.
Where the gap actually sits
The steelman for tighter rules is real. Biometric data is not a password — it cannot be reset after a breach, and Beijing lawyer Yile Deng's warning to Rest of World is accurate: "once a person's facial data enters an AI marketplace, users may effectively lose long-term control over how their biometric identity is used." A face licensed for one 44-episode drama can be re-synthesized indefinitely, resold, or stripped of the metadata that ties it back to a consent record. That is precisely the harm PIPL's separate-consent requirement was meant to prevent, and the scale of unauthorized use now surfacing shows the harm is not hypothetical: ByteDance's short-drama app Hongguo alone removed more than 85,000 videos involving unauthorized AI reproduction of people's faces and voices since the start of 2026, and the Guangzhou Internet Court has heard roughly 700 AI-related face-theft cases over the past three years (Rest of World).
Those two numbers describe different problems. The 85,000 takedowns are Hongguo's own platform-level enforcement working — a private company policing unauthorized use of faces that were never licensed in the first place. The 700 court cases are what happens when platform enforcement fails and a victim has to litigate individually, as one actress did in March 2026, when the Beijing Internet Court ruled that a production company and streaming platform had violated her image rights by face-swapping her likeness into a drama without authorization — even though the AI-generated face was not a perfect match. Judge Zhao Qi's finding that a face is protected "if recognizable by the public," regardless of pixel-level fidelity, set a low bar for what counts as unauthorized use (Beijing Internet Court).
That is the real gap: not an absence of consent rules, but an enforcement model that relies on courts adjudicating cases one at a time against a marketplace generating licensed faces at commercial scale. PIPL and the Deep Synthesis Provisions put the consent obligation on the party doing the editing — the studio or app rendering a new face. They say less about what obligations attach to the marketplace platforms sitting upstream, aggregating and reselling biometric catalogs for third-party use. ActID and New Claw are not obviously "editing" anyone's face; they are warehousing and licensing it, a function PIPL's sensitive-information rules cover in principle but that no regulator has yet built a specific compliance checklist for — verified consent chain-of-custody, resale caps, or mandatory registration of biometric marketplaces the way the CAC already requires algorithm registration for deep-synthesis providers.
The fix China needs is not a new law restricting the face-licensing market — that would foreclose a legitimate income channel for the people using it and push the same demand toward unregulated, unpaid scraping instead. It is extending the deep-synthesis registration and consent-verification model that already applies to editing platforms upstream, to the licensing marketplaces that supply them. Proportionate regulation here means making the existing consent regime enforceable at the platform level before litigation becomes the primary check on a market moving far faster than 700 lawsuits can keep pace with.