Our Insights

Eric Johnson 7/13/26 Eric Johnson 7/13/26

New Draft Rules for Digital Avatars and Virtual Humans in China

On April 3, 2026, China’s Cyberspace Administration (CAC) released draft Measures for the Management of Digital Virtual Human Information Services (《数字虚拟人信息服务管理办法（征求意见稿）》), for public comment through May 6. These measures, if finalized, would regulate the use of “digital virtual humans” for a variety of commercial purposes online, including advertising. In effect, the draft extends and elaborates existing PRC rules on deep synthesis, AI-generated content, personal information protection and online information services, to the realm of digital avatars.

For the advertising industry, this is notable because the draft is not aimed only at AI developers. It appears designed to capture the broader commercial chain: brands, agencies, networks, service providers and platforms may all have responsibilities depending on their role.

Why this matters for marketing teams

Digital avatars are increasingly used for livestreaming, customer contact, endorsements, and multilingual campaigns. China’s draft rules suggest regulators now view these tools as mainstream commercial media to be regulated accordingly, including through the application of existing advertising, consumer protection and data rules.

Key themes in the draft

1. Clear labeling of virtual humans

The draft requires prominent disclosure of any use of a digital human. This aligns with current AI labeling rules, and will likely apply throughout the use, not just in fine print or at the start of a video.

2. Consent and likeness rights

Use of a real person’s face, voice, image or identifiable traits may require consent. This overlaps with China’s existing personal information and personality-rights regimes, and we would expect it to be particularly relevant for “lookalike” spokespeople or scraped training materials tied to individuals.

3. False or manipulative promotion

The draft prohibits using digital humans for false promotion or improper inducement to consume goods or services. This is already prohibited under existing Chinese law. The inclusion in the draft indicates that false advertising by virtual human will not be treated any differently.

4.Regulatory compliance obligations

Service providers offering virtual human products “with public opinion or social mobilization capabilities” must complete an algorithm filing, make timely updates for any changes, and conduct mandatory security assessments. Platforms are also required to establish convenient appeal, complaint and reporting mechanisms to handle user feedback on virtual human products deployed on their platforms.

5. Penalties for violations:

The draft proposespenalties such as warnings, public reprimands and mandatory corrections, service suspensions, and fines of up to RMB 200,000. More serious penalties may be imposed if provided for under other relevant laws, which could plausibly include China’s very powerful telecom laws as well as consumer protection laws. In general, the draft reflects China’s multi-agency approach to AI governance, potentially involving advertising, telecom, public security and sector-specific regulators, depending on the situation.

Practical steps for brands now

The draft remains open for comment, but the direction is clear. China is moving toward treating virtual-human marketing as an ordinary regulated commercial activity, with potentially broad liability, and not as an experimental or niche AI tool. Brands and agencies considering the use of avatars for marketing in China should closely watch the development of these proposed measures, and in the meantime review any existing virtual avatar campaigns for prospective compliance. Looking forward, it is likely that compliance costs for these tools will rise due to consent, labeling, filing and review obligations.

By Eric Johnson

Eric Johnson 7/13/26 Eric Johnson 7/13/26

Regulating AI Marketing in China – GEO as the Starting Point

A recent exposé on a state broadcaster in China called out abuses in the “Generative Engine Optimization” (GEO) market, highlighting fresh regulator attention to the mass production of synthetic promotional content designed to influence AI chatbot outputs and recommendation systems.

3.15 Exposure – the Policy Signal

CCTV’s annual March 15 "Consumer Rights Gala” program exposed what was described as an “AI poisoning” industry, in which firms flood the internet with fabricated reviews, product claims and endorsements in order to manipulate AI-generated recommendations and consumer decisions. Following the broadcast, regulators and industry experts warned that data poisoning of AI models could become a significant consumer-protection risk.

Historically, this program has functioned as a policy signal, often preceding enforcement campaigns or new regulations in consumer-facing sectors featured on the program.

Legal Framework - Leveraging existing laws for now

To date, most authorities and commentators have concluded that GEO-like practices fall within the scope of existing laws, including:

The Advertising Law, which prohibits false or misleading commercial claims, and requires clear identification of advertising content;
The Anti-Unfair Competition Law, which addresses fabricated transactions, reputation manipulation and other conduct that distorts competitive conditions and market order; and
The Consumer Rights Protection Law, which protects consumer rights to truthful information and fair transactions.

AI-specific regulations are also relevant, such as:

The Interim Measures for the Management of Generative AI Services (2023), which requires public-facing generative-AI providers to conduct security assessments, register recommendation algorithms and ensure lawful, safe and truthful content outputs; and
The Administrative Provisions on Algorithm Recommendation for Internet Information Services (2022), which also require filing and risk-management measures for recommendation systems with significant social or public-opinion impact.

Taken together, these law could easily be construed to prohibit common GEO practices, like the mass generation of fabricated promotional content, manipulation of algorithmic rankings or recommendations, and failure by platforms or AI providers to implement adequate content risk controls.

Historical Comparison: 3.15 Exposure of digital advertising / DSP

A decade ago, targeted advertising faced a similar exposé on the March 15 program, highlighting illegal collection and trafficking of user data, traffic fraud, malicious targeting, misleading placement and attribution, etc. Directly following this program and for the next several years (roughly 2016-2019) the digital advertising ecosystem came under intense scrutiny, including a substantial revision to the Advertising Law to directly address unique digital advertising issues, and several enforcement campaigns from the CAC and SAIC / SAMR. We would expect similar developments AI-related advertising now.

Post‑3.15 rapid regulatory and legislative action

Already, comprehensive national AI regulation has been re-prioritized. In early March, the 2026 Two Sessions (China’s annual national political and legislative agenda-setting meetings) explicitly prioritized comprehensive AI legislation, with GEO data poisoning, AI content authenticity, and platform review obligations as core topics. Key proposals included (i) re-starting work on a national AI law, (ii) clarifying liability for developers, platforms, and users, and (iii) mandating AI content watermarking and traceability.

In addition, early in 2026, the MIIT’s China Academy of Information and Communications Technology (CAICT) held standard-setting seminars and working groups on AI safety and governance with more than 20 industry, academic, and research institutions, including discussions of fake data injection and forged information sources, compliance baselines for content authenticity, traceability and transparency. Proposals included launching a first round of GEO service credibility evaluations in Q1 2026, and adding dedicated clauses to the Anti‑Unfair Competition Law to address AI data poisoning and deceptive marketing.

Enforcement trajectory and regulatory signaling

While the broadcast did not explicitly propose any new laws, it foreshadows heightened scrutiny and legal action relating to GEO issues which is already beginning to take shape. This creates a newly urgent need for risk analysis for the many parties potentially touched by GEO, which would include practitioners and purchases of GEO services, and both the platforms and brands potentially subject to GEO manipulation.

In that context, for businesses operating in China’s digital marketing and AI ecosystems, these developments give strong grounds for continued vigilance around content authenticity and truth-in-advertising, including:

Not using synthetic review-generation or ranking-manipulation services;
Enhancing lear labeling and verification protocols for all advertising and AI-generated content;
Revisiting the always-evolving compliance obligations for advertisers and ad platforms, especially when operating digitally; and
Continuing to monitor enforcement patterns and regulatory signals, in anticipation of more frequent investigations and targeted rule-making.

By Eric Johnson

Eric Johnson 3/15/25 Eric Johnson 3/15/25

The Shanghai Whitelist Pilot for Data Transfers

Cross-border data transfers (“CBDTs") have become a critical compliance concern for multinational companies in China, including Chinese companies with a global footprint. To address some of the inefficiencies that CBDT compliance requirements have introduced, China is now trialing regional deregulation in free trade zones (“FTZs”). One of the more notable recent developments is the testing of a whitelist approach by the Shanghai FTZ, similar to the whitelist approach seen under the GDPR.

On May 17, 2024, the Shanghai FTZ Lingang Special Area released the first three trial lists (collectively, the “Lingang Whitelists”) identifying data types that will be subject to reduced cross-border data transfer restrictions. The lists apply to the data generated in three specific industries, as follows:

Observations on the EU Artificial Intelligence Act and related Chinese legislation

2023.05.30

一、欧盟人工智能法案概览

从 ChatGPT 的“惊艳登场”，到微软结合自身优势最新推出的 Windows 11 Copilot，人工智能（“AI”）技术以及相关应用在全球范围内发展和部署可谓突飞猛进。在 AI 浪潮给经济和生活注入活力的同时，各国/地区监管者均不同程度上表达了对 AI 技术及相关应用的担忧，并拟通过政策、立法等对 AI 进行监管。其中，备受关注的无异于 2021 年 4 月 21 日，欧盟委员会（European Commission）发布的欧盟人工智能法案草案（“ 欧盟 AI 法案 ”或“ 法案 ”）1 。从立法进展看，2022 年 12 月 6 日，欧盟理事会（European Council）通过了关于欧盟 AI 法案的共同立场（Common Posi)on）版本。2023 年 5 月 11 日，欧洲议会（European Parliament）的内部市场委员会（Internal Market CommiKee）和公民自由委员会（Civil Liber)es CommiKee）以 84 票赞成、7 票反对和 12 票弃权的形式通过了欧盟 AI 法案的谈判授权草案2 。整体时间表上，欧盟 AI 法案有望于 2023 年第四季度通过。下表为我们对法案相关要点的简要总结3 。

See more here.