1. What is the current legal landscape for Artificial Intelligence in your jurisdiction?

China has not yet enacted a comprehensive, dedicated artificial intelligence (“AI”) law. Instead, it has developed a multi-layered AI governance framework characterised by a “small-incision” legislative approach, under which regulators introduce targeted rules addressing specific technologies, application scenarios and emerging risks, while relying on the existing legal framework governing cybersecurity, data protection and information regulation.

At the statutory level, China has already released the Cybersecurity Law (“CSL”), the Data Security Law (“DSL”) and the Personal Information Protection Law (“PIPL”), which govern cybersecurity and network activities, data of national importance and personal information. As these laws establish baseline requirements regarding data and cybersecurity, the laws apply to AI systems throughout their development, deployment and operation. Depending on the nature of the AI service, additional sector-specific regulations may also apply.

Building upon these laws, China has introduced a series of administrative regulations targeting different aspects of AI technologies. These regulations address issues including recommendation algorithms, deep synthesis technologies, generative AI services, AI-generated content labelling, AI ethics governance and anthropomorphic AI services.

• The Provisions on the Administration of Algorithmic Recommendation in Internet Information Services (“Algorithm Recommendation Provisions”) establish a filing and security assessment regime for algorithm recommendation services possessing public opinion attributes or social mobilisation capabilities.
• The Administrative Provisions on Deep Synthesis in Internet-based Information Services (“Deep Synthesis Provisions”) regulate deep synthesis technologies through requirements on identity verification, content management, synthetic content identification, training data governance and security assessment.
• The Interim Measures for the Management of Generative Artificial Intelligence Services (“Gen AI Measures”) apply to providers offering generative AI services to the Chinese public. Key requirements include completing a security assessment (for certain service providers), use of lawful training data, protection of personal information, prevention of discriminatory or unlawful content, implementation of complaint and correction mechanisms, and safeguards for minors.
• The Measures for the Labelling of AI-Generated and Synthesised Content introduce a dual-labelling mechanism consisting of explicit labels visible to users and implicit identifiers embedded in metadata. Different labelling requirements apply to text, images, audio, video and virtual scenes.
• The Measures for the Administration of Artificial Intelligence Technology Ethics Review and Services (Trial) establish China’s first nationwide framework for AI ethics review, which adopts a two-tier review mechanism.
• The Interim Measures for the Administration of Anthropomorphic Artificial Intelligence Interaction Services (“Anthropomorphic AI Interaction Measures”) regulate AI companion services that simulate human personalities or provide sustained emotional interaction, and introduce requirements relating to user protection, content governance, protection of minors, addiction prevention and security management.
• The Draft Measures for the Administration of Digital Virtual Human Information Services propose a dedicated regulatory framework for digital virtual human services. Although still under consultation, these draft measures indicate that future regulation is likely to address issues including personality rights, personal information protection, identity disclosure and risk management throughout the lifecycle of digital virtual human services.

In addition to legislation and administrative regulations, technical standards constitute an important component of China’s AI governance framework. Although most technical standards are voluntary rather than legally binding, they provide detailed implementation guidance for regulatory requirements and increasingly serve as important compliance benchmarks in regulatory enforcement and industry practice.

2. What three essential pieces of advice would you give to clients involved in Artificial Intelligence matters?

A. Role Classification and Compliance Mapping

For enterprises entering the AI space, a critical first step is to accurately define their legal roles based on the chosen AI deployment scenarios. Market players typically adopt a range of approaches, including developing proprietary large language models (“LLMs”), leveraging open-source frameworks, and integrating third-party LLMs through APIs. These distinct operational approaches directly shape an entity’s regulatory classification—such as a service provider, technical supporter, or end-user—each triggering a standalone set of compliance obligations.

From a regulatory enforcement perspective, service providers bear comprehensive liabilities. Their compliance obligations span content security reviews, IP risk mitigation, data privacy compliance, and the completion of both algorithm and generative AI service registration (“Dual Filing”). Moving up the supply chain, technical supporters enjoy lighter compliance obligations but must still discharge core responsibilities regarding content vetting, data protection and algorithm filings. For downstream corporate end-users, the compliance focus shifts to managing contractual obligations and the preservation of critical corporate data assets. Consequently, a granular mapping of these legal roles enables enterprises to construct tailored, stage-by-stage compliance checklists, allocating responsibilities between internal stakeholders and third-party LLM providers. Such mapping and compliance checklists should cover several factors, including:

• Vendor Integration: Conducting rigorous due diligence on upstream LLM providers (focusing on the status of their Dual Filing duties, server hosting locations and data security infrastructure) and executing comprehensive service agreements that clearly delineate corporate data privacy, output ownership and IP allocations.
• Model Training: Implementing a robust data governance framework to review the legality and validity of training datasets, while embedding proactive internal controls for algorithmic governance and risk mitigation.
• Pre-Market Launch: Running a comprehensive compliance audit to clear necessary regulatory hurdles, ensuring completion of the Dual Filing steps, content watermarking requirements, platform moderation protocols and user grievance channels.

B. Determining Dual Filing Obligations

Prior to the commercial launch of AI products or the public provision of AI services, a fundamental compliance threshold for enterprises is assessing whether they fall within the Dual Filing framework. The regulatory scope of algorithm registration and generative AI service filing is determined by distinct statutory triggers:

• Algorithm Filing: The filing requirement is triggered when an enterprise deploys any of the five statutory categories of algorithmic capabilities: synthetic generation, personalised recommendation, ranking and curation, search and filtering, or scheduling and decision-making. The obligation is triggered by the technical nature of the algorithm—irrespective of whether the enterprise acts as a service provider or a technical supporter, and regardless of whether the user base is B2B (enterprise) or B2C (individual).
• Generative AI Service Filing/ Registration: This requirement is triggered by the content and potential societal impact of the AI service. Enterprises must assess whether their public-facing AI services involve content generation functions that may influence public opinion or facilitate social mobilisation. Where such attributes are present, a formal service filing with the regulator is required. However, where an enterprise merely integrates a third-party foundational LLM that has already successfully completed its generative AI service filing (e.g., through an API), without conducting additional corpus training or model fine-tuning, the enterprise is exempted from the full filing procedure. Instead, the enterprise is only required to complete a generative AI service registration, which involves a more streamlined and expedited regulatory procedure compared to the full filing track.

C. Content Moderation and Labelling Mandates

Under the current Chinese regulatory framework, AI service providers are recognised as online information content producers and bear primary accountability for network information security. This regulatory intersection creates significant compliance obligations and is subject to heightened scrutiny. Enterprises should implement the following operational safeguards:

• Proactive Content Filtering: Establish robust, real-time filtering and blocking mechanisms to intercept content explicitly prohibited or restricted under applicable laws.
• Content Safety Controls: Guarantee the integrity of AI-generated content by systematically preventing the output of illicit or harmful information. This includes material that endangers national security, spreads rumours, or involves pornography, gambling, violence or defamation.
• Establish a Prompt Complaint Handling Mechanism: Implement procedures to ensure the timely recording, investigation and resolution of user-submitted complaints and enforcement reports, with appropriate notification of remedial measures.

To insulate enterprises from liabilities associated with deepfakes or public misinformation, AI service providers must establish a compliant content-labelling framework. Under the current regulatory framework, such obligations follow a dual-track approach:

• Explicit Labelling: Designed to promote consumer transparency by identifying whether content is AI-generated, this requirement obliges providers to apply visible labels to AI-generated content or live user interfaces, thereby fulfilling the regulatory intent of public disclosure and reducing the risk of confusion.
• Implicit Labelling: Designed to enable content traceability by recording the origin and dissemination history of AI-generated content, this requirement involves embedding invisible technical markers within the file’s metadata or underlying infrastructure. These markers support regulatory oversight by capturing key information, including semantic attributes, service provider codes and content identifiers, to ensure downstream accountability.

3. What are the greatest threats and opportunities in Artificial Intelligence law in the next 12 months?

A. Greatest Opportunities

China’s AI industry is expected to maintain robust growth over the next 12 months, generating significant new legal demands and creating substantial opportunities for the development of AI-related legal services. According to official statistics, the scale of China’s AI industry exceeded one trillion yuan in 2025, with a projected annual growth rate of over 30% in 2026. The rapid expansion of the AI industry will inevitably drive surging market demand for professional legal services across the sector.

China has continuously rolled out supportive policies to fuel high-quality AI development. In June 2026, the China Securities Regulatory Commission (CSRC) officially expanded the Fifth Listing Standard of the Sci-Tech Innovation Board to cover the AI sector, actively supporting high-quality large AI model enterprises to pursue public listings. This policy marks the formal implementation of tailored listing rules for AI firms on the Sci-Tech Innovation Board, providing more accessible financing channels for AI enterprises with strong technological capabilities, including those at the early stages of commercialisation. The continued development of the AI capital market is expected to generate increasingly diverse AI-focused legal service needs, including investment and financing, IPO compliance, algorithm filing, generative AI service registration and risk assessment.

The accelerated development of China’s AI governance framework is also expected to generate further demand. On May 8, 2026, the State Council’s 2026 Legislative Work Plan explicitly proposed improving AI governance and advancing comprehensive legislation to support the sound development of AI. Meanwhile, the Standing Committee of the National People’s Congress has included legislative initiatives for the healthy development of AI in its 2026 preliminary legislative agenda. The next 12 months are expected to be an important period for the further development of China’s AI governance framework, with new rules likely to create ongoing demand for specialised legal services.

B.Core Threats

a) The Existing Regulatory System Fails to Resolve Fundamental Industrial Bottlenecks

China has established a multi-layered AI governance framework through a flexible, iterative, small-incision legislative approach, covering key administrative rules such as the Algorithm Recommendation Provisions, Deep Synthesis Provisions, and Gen AI Measures. However, departmental regulatory measures alone are insufficient to address numerous structural problems constraining industrial development.

The governance of AI training data remains the most prominent institutional bottleneck. No overarching legal framework has been established to clarify the rights relating to the ownership, data circulation and profit distribution mechanisms of training datasets. The strong demand from AI developers for the large-scale data used in model training creates significant tension with the interests of data rights holders in protecting their data from unauthorised use, leading to a growing number of disputes and industry controversies.

In addition, the liability allocation framework across the AI value chain remains ambiguous. Current laws do not clearly divide legal responsibilities among foundation model developers, application-layer deployers and end users. As emerging technologies, including AI Agents and embodied intelligence, are deployed from laboratory scenarios to real-world environments—enabling autonomous operation, digital system manipulation and even control of physical equipment—the absence of a definitive liability framework exposes all industrial participants to unpredictable legal risks and uncertain liability.

b) Emerging AI Application Scenarios Outpace Regulatory Development

Technological iteration continues to outpace the formulation and implementation of regulatory rules. For instance, the Anthropomorphic AI Interaction Measures, which took effect on July 15, 2026, is China’s first dedicated regulation targeting anthropomorphic emotional interactive AI systems. Nevertheless, it only covers a segmented field of the rapidly expanding AI industry. Regulatory frameworks for emerging formats such as AI Agents, AI-native applications and embodied intelligence remain largely underdeveloped.

During this regulatory gap period, relevant enterprises can only conduct compliance self-assessments based on the general principles of the CSL, DSL and PIPL, resulting in limited legal certainty and substantial compliance ambiguities for innovative AI businesses.

4. How do you ensure high client satisfaction levels are maintained by your practice?

Our client base spans a diverse spectrum, ranging from early-stage startups to large-cap listed companies, with many of these relationships extending well beyond a decade. Despite differences in industry sector, market capitalisation and stage of development, we have consistently observed that clients’ core expectations of external legal counsel remain broadly consistent: a precise, timely and forward-looking interpretation of legal rules, coupled with a profound understanding of commercial practices within their respective industries.

Based on this understanding, we enhance client satisfaction through two principal approaches:

• First, we maintain deep engagement in legal research and legislative support. Rather than passively reacting to the release of new laws and regulations, we proactively invest in the study of frontier legal issues and participate in policy discussions. Our partners and senior lawyers are regularly invited to contribute to legislative research, judicial interpretation consultations and the formulation of industry-specific regulatory frameworks. This upstream involvement enables us to anticipate regulatory directions and advise clients on strategic positioning well before rules are finalised, helping them strike a balance between compliance imperatives and business objectives. When clients navigate complex legal terrain, we provide not merely doctrinal analysis of statutory provisions, but in-depth assessments grounded in the underlying legislative intent and policy rationale.
• Second, we support clients through lawyer secondments and conduct ongoing industry-focused research. We routinely assign lawyers to work within our clients’ organisations, whether through on-site secondments or project-based placements, enabling them to gain first-hand insights into the clients’ day-to-day operations and major transactions. This embedded service model affords us first-hand exposure to the real-world mechanics of our clients’ industries and a granular appreciation of their commercial decision-making logic and pain points. Complementing this, we maintain a structured industry research programme that continuously tracks market developments, regulatory trends and leading cases across our key practice areas. This fusion of legal acumen with sector-specific know-how allows us to deliver actionable solutions which are not only legally sound but also commercially pragmatic.

Ultimately, client satisfaction is built upon value creation. We believe that long-term trust is sustained only when clients recognise that their legal team truly understands their business and is capable of delivering proactive, forward-looking judgment that exceeds expectations.

5. What technological advancements are reshaping Artificial Intelligence law and how can clients benefit from them?

Generative AI stands as one of the most disruptive technological breakthroughs in recent years, fundamentally reshaping how society operates and where the boundaries of legal rules are drawn. From content creation to decision support, and automated services to personalised recommendations, generative AI has permeated nearly every industry. This wave of technological change not only brings leaps in efficiency but also triggers a profound redistribution of interests—the relationships between creators and platforms, data providers and model developers, and algorithm users and affected parties are all being redefined. In response, legislators, regulators and judiciaries across the globe are accelerating the construction of new legal frameworks to strike a balance between incentivising innovation and mitigating risk.

Against this backdrop, we have carefully examined the technical characteristics of generative AI and considered how to embed it effectively into our legal service delivery to create tangible value for clients. Our core assessment is that legal tasks characterised by two features—a sufficient data foundation and verifiable output—are particularly suitable for AI-assisted efficiency improvements, ultimately enabling clients to benefit on both cost and time savings.

Specifically, clients can draw on mature AI-assisted solutions in the following areas. The first is bulk contract review. When clients face hundreds or thousands of transaction documents with highly similar clauses and structures, traditional manual line-by-line review is not only time-consuming but also prone to oversight due to fatigue. With AI-assisted tools and professional oversight, clients can significantly accelerate first-pass reviews, enabling them to complete compliance checks at lower cost and within shorter timeframes. The second is large-scale case and regulatory research. For projects requiring cross-jurisdictional searches through vast numbers of precedents and regulatory filings, AI can rapidly aggregate information and perform preliminary analysis, enabling lawyers to concentrate their efforts on strategic judgment rather than repetitive retrieval tasks.

We have always maintained that technology itself is not the end goal; rather, its purpose is to enable the delivery of higher-quality legal outcomes for clients. The true value of generative AI lies in reducing lawyers’ reliance on repetitive tasks, allowing them to devote greater attention to understanding clients’ commercial objectives, anticipating legal risks and providing forward-looking advice. Clients who embrace this technological transformation will be better positioned to access more efficient and precise legal support in an increasingly complex business environment.