Original AI law research from Geason: algorithm filing, data compliance, global expansion, IP, compute and capital governance, with usable conclusions.
An option pool and a bank facility are usually negotiated by different people at different times, yet both run through the same cap table. This article sets out how Chinese AI companies size and paper an ESOP, when the tax deferral under Caishui [2016] No. 101 applies and where it fails, what lenders actually diligence in a debt financing, and the three points where the two instruments collide.
An AI company that hires a engineer in another country, or sends one abroad for six months, creates two exposures at once: an employment-law question about who is the lawful employer, and a tax question about whether the company has just acquired a taxable presence. This article covers employing entities, EOR and secondment structures, work permits and social insurance, permanent establishment risk from engineers and from API sales, transfer pricing and revenue characterisation.
Deploying a model on an overseas server usually triggers two compliance tracks at once: technology export control and cross-border data transfer. They have different objects, legal bases, filing routes and timelines, and neither substitutes for the other. This article sets out a comparison table and four common scenarios.
The comment period on Announcement No. 48 closes on 15 November 2026. Today exporters file a pre-export registration; once the revised catalogue takes effect, the same technology may fall into the prohibited or restricted category, and the procedure shifts from information filing to licensing. The transition window is where companies can still shape the outcome.
On 16 October 2026, China's Ministry of Commerce and Ministry of Science and Technology issued Announcement No. 48, proposing to add 40 technologies across AI, biopharma and solar to the Catalogue of Technologies Prohibited or Restricted from Export. Twenty-three are AI-related. The catalogue is still out for comment, but the pre-export registration duty took effect on publication day.
On 31 August 2026, the General Office of the Ministry of Industry and Information Technology issued the Notice on Carrying Out the Special Action to Cultivate AI Application Service Providers (MIIT General Office Science and Technology Letter [2026] No. 414). The Notice defines the scope of 'AI application service providers' for the first time, sets targets of more than 2,000 providers in the resource pool by the end of 2026 and no fewer than 3,000 by the end of 2027, and arranges four tasks: building the resource pool, improving service supply, promoting large-scale application and strengthening support. This article reviews the positioning, tasks and compliance requirements relevant to AI companies.
On 21 August 2026, China's cyberspace portal published the Action Plan for Promoting the High-Quality Development of Cyberspace Enterprises (2026-2030). Built around cultivating and serving quality cyberspace enterprises, the plan sets out five actions — tiered cultivation, innovation capability, industrial upgrading, overseas market expansion and safe and orderly development — signalling policy support for AI key-technology research and overseas compliance services, while setting out regulatory requirements on comprehensive algorithm governance and preventing disorderly competition. This article reviews the provisions relevant to AI companies.
On 20 August 2026, the Measures for Network Data Security Risk Assessment (Order No. 24 of three departments), jointly issued by the Cyberspace Administration of China, the Ministry of Industry and Information Technology and the Ministry of Public Security, took effect. The Measures establish an annual risk-assessment duty for important data processors, a reporting and verification mechanism for assessment reports, and independence requirements for third-party assessment institutions. This article reviews the assessment subject, frequency, reporting and key compliance points, and their implications for AI companies.
The Measures for the Security Management of the Application of Face-Recognition Technology (Order No. 19 of the Cyberspace Administration of China and the Ministry of Public Security) took effect on 1 June 2025. The Measures set the filing trigger at the storage of face information of 100,000 people, require filing with the cyberspace authority at or above the provincial level within 30 working days, and require five categories of documents. This article reviews the filing trigger, document list, the non-mandatory principle and processing rules, and their implications for AI companies.
On 7 August 2026 the Cyberspace Administration of China opened public consultation on the Provisions on Personal Information Protection by Large Personal Information Processors (Draft for Comment), with comments due by 7 September. The draft consolidates two earlier drafts on 'large online platforms' and sets dedicated obligations for operators processing the personal information of more than 10 million people: proactive filing and designation, appointment of a personal information protection officer, an annual social responsibility report, and filing of impact assessment reports. This article reviews the designation criteria, core obligations and implications for AI companies.
On 31 July 2026 the European Commission announced that the AI Act would expand its scope from 2 August, applying the transparency rules and the rules for providers of general-purpose AI models. Interactive AI such as chatbots must tell users they are interacting with AI; deepfake content must be labelled and carry machine-readable markings; and obligations for general-purpose AI model providers take effect in parallel. This article reviews the application milestones, core obligations and implications for Chinese companies going global.
In July 2026 the National Data Administration expressly encouraged business-model innovation based on token applications, exploring new trading models such as token trading, cultivating a market consensus of 'paying for quality data', and deepening the integration of token applications with industry, healthcare and finance. This article observes the contract, pricing, data and tax issues that the token economy raises from a compliance perspective.
On 28 July 2026 the National Data Administration disclosed at a press conference for the 2026 China International Big Data Industry Expo that 72 industry chain-leader units had built more than 45 PB of high-quality datasets covering over 759 high-value scenarios, and that 1.4 million PFLOPS of intelligent computing power had been connected to the national integrated computing network trial platform. This article reviews progress on high-quality datasets and computing infrastructure and what it means for the supply side of AI companies.
On 24 July 2026 the Cyberspace Administration of China issued the Provisions on Simplified Measures for Personal Information Protection by Small Personal Information Processors. The Provisions set streamlined measures for small processors on publishing processing rules, notice and consent, rights response and deletion, and clarify the conditions under which they are exempt from filing a cross-border data security assessment, concluding standard contracts or obtaining certification when providing personal information overseas. This article reviews the identification of applicable entities and the boundaries of the streamlined duties.
In July 2026 the Cyberspace Administration of China deployed a four-month 'Qinglang — Minors Online Protection' special action, whose first phase is the 2026 summer campaign on the online environment for minors. The action focuses on AI 'poison animation' and the hyping of child influencers. This article reviews the regulatory red lines drawn by the special action and what they mean for companies providing AI content and interaction services.
From 17 to 20 July 2026 the 2026 World AI Conference and High-Level Meeting on Global AI Governance was held in Shanghai. During the conference the World AI Cooperation Organisation was established with its headquarters in Shanghai, and the International AI Ethics Governance Action Plan, the AI Cooperation and Development Action Plan and the Global Cooperation Initiative on Agent Interoperability were released. This article reviews what these outcomes mean for the cross-border compliance of AI companies going global.
The Interim Measures for the Administration of Anthropomorphic Interactive AI Services (Order No. 21 of five departments) took effect on 15 July 2026 — China's first departmental rules dedicated to emotionally companionable AI. The Measures define the scope of anthropomorphic interactive services, set out six categories of prohibited generated content, lay down special rules for minors, and specify when a security assessment is triggered. This article sets out the scope of application, core obligations and compliance arrangements from an enterprise perspective.
On 15 July 2026 the Cyberspace Administration of China announced seven newly filed on-device generative AI services for mobile handsets, including Apple Intelligence. On-device large-model filing has entered a normalised phase. This article reviews how on-device filing differs from cloud filing, the division of duties between model providers and device makers, and the compliance points for on-device products on data localisation, content labelling and minors protection.
On 8 July 2026 the Ministry of Human Resources and Social Security, the National Development and Reform Commission, the Ministry of Industry and Information Technology and the National Data Administration jointly issued the Opinions on Accelerating the Development of 'AI + Human Resources and Social Security' Applications. The document calls for building high-quality industry datasets and AI corpora for the human resources and social security sector, promoting AI in employment, social insurance and labour-relations scenarios. This article reviews the industry-dataset requirements and compliance boundaries relevant to AI companies.
On 6 July 2026 the Cyberspace Administration of China reported on the first phase of the 'Qinglang — Rectifying AI Application Disorder' special action. Launched in April 2026, the action focuses on failures to perform large-model filing and registration duties, inadequate platform security and content-filtering capability, AI data poisoning, and incomplete implementation of generative-content labelling. This article reviews the four problem categories identified in phase one and what they mean for companies' compliance work.
On 3 July 2026 the Cyberspace Administration of China re-opened public consultation on the revised draft of the Measures for the Administration of Internet Information Services. The administrative regulation, in force since 2000, is being systematically revised, with two new dedicated sections — 'platform information services' and 'intelligent information services'. This article sets out the institutional points of the intelligent information services section — generative-content labelling, constraints on algorithmic recommendation, protection of workers' rights, and responses to significant security risks — and explains what it means for companies providing AI information services.
In July 2026 the Ministry of Industry and Information Technology launched its 2026 open call for innovation tasks in the industry and information technology field. Under the artificial intelligence heading it set four tracks — foundational technology, application-enabling technology, intelligent products and equipment, and common supporting technology — while the future-industries heading covers five tracks including humanoid robots and embodied intelligence, and brain-inspired intelligence. This article reviews the task layout, the logic of the open-call mechanism, and the compliance groundwork applicants should prepare on intellectual property, data and ethics governance.
On 30 June 2026 Meituan released LongCat-2.0, a MoE model with 1.6 trillion total parameters and the first trillion-parameter model in China trained and run end to end on domestic compute, alongside an open-source programme. This article views the event through open-source compliance, model filing and downstream responsibility.
On 29 June 2026 the State Council executive meeting heard a report on AI development and made a series of targeted deployments. It stressed both accelerating AI breakthroughs and implementing the 'AI Plus' initiative, and safeguarding AI security with ethics and testing-certification rules. This article reviews the signals and their meaning for companies.
On 26 June 2026 the State Administration for Market Regulation approved the release of seven national standards on 'AI — Agent Interconnection'. As AI moves from perception and understanding into generation and decision-making, interconnection between agents becomes a foundational industry issue. This article reviews the standards and the new compliance questions raised by agent scale-up.
In June 2026 the Ministry of Transport, together with several national railway, civil aviation and postal bodies, released an action plan on innovative application scenarios for 'AI Plus Transportation', setting out ten directions and an initial 41 priority scenarios. This article reviews the plan and the data and liability issues in scenarios such as intelligent driving and shipping.
On 18 June 2026 eight departments including the Ministry of Commerce released an implementation opinion on accelerating 'AI Plus Consumption', with 17 measures across five areas, expanding smart-terminal supply and enabling service consumption. This article reviews the policy and the compliance points around smart terminals and consumer scenarios.
On 18 June 2026 the National Financial Regulatory Administration issued guidance on the secure development and use of AI in banking and insurance, requiring financial institutions to build a full-lifecycle management system, strengthen access management for high-risk applications, and reinforce data security and personal information protection.
To support the 'Qinglang — Rectifying AI Application Disorder' special action, the CAC Reporting Centre opened a dedicated reporting channel for 'AI application disorder' from June 2026. The channel changes where enforcement leads come from. This article reviews what that means for companies' compliance work.
On 8 June 2026 the National Data Administration released the Implementation Plan for the Initiative to Promote the Construction of High-Quality Industrial Datasets, setting out six special actions. This is China's first systemic national-level deployment on how data empowers AI development. This article reviews the plan and the data-compliance prerequisites dataset building cannot avoid.
On 2 June 2026 the US President signed an executive order strengthening AI model review, with the pre-release review window reportedly cut from 90 days to 30 days. Combined with a later shift of AI safety incident reporting from voluntary to mandatory, this article reviews the direction of US AI regulation and what it means for Chinese AI companies entering the US market.
On 3 June 2026 the Cyberspace Administration of China set out the thinking behind its 'small, fast, flexible' approach to regulating anthropomorphic AI interaction services. This article reviews the regulatory logic behind choosing dedicated legislation over general rules, and what it means for product design and compliance systems.
On 29 May 2026 the Cyberspace Administration of China issued a notice launching a two-month 'Qinglang — Optimising the Business Online Environment and Rectifying Malicious Speculation about Enterprises' special action. It targets four categories of prominent problem and, for the first time, expressly lists AI-generated false corporate information. This article reviews the action and what it means for AI content producers and companies.
On 29 May 2026 the Cyberspace Administration of China, the Ministry of Education, the Ministry of Industry and Information Technology and the Ministry of Human Resources and Social Security jointly issued the 2026 Work Points on Improving Digital Literacy and Skills for All, deploying six areas and fifteen priority tasks. Two lines relate directly to the AI sector: improving AI literacy for all, and promoting the safe and standardised development of AI. This article reviews the document and its implications.
From 28 to 31 May 2026 the World Intelligent Industry Expo, co-hosted by the Tianjin and Chongqing municipal governments, was held in Tianjin under the theme 'Intelligence Moves the World, Empowering the Future'. Platform launches and project signings were dense. This article reads the event for signals on industrial deployment, cross-border compliance and AI governance.
On 27 May 2026 the Equipment Industry Development Centre of the Ministry of Industry and Information Technology released the 2026 automotive standardisation work points. Under 'standards-led action in emerging fields', it calls for accelerating intelligent connected vehicle standards, including mandatory national standards for autonomous driving systems and simulation test methods. This article reviews the points and their meaning for liability boundaries and data compliance.
In May 2026 the National Data Administration issued its 2026 Digital Economy Work Points, setting out eight areas of priority tasks. Two lines matter most for companies: deepening the market-based allocation of data elements, and strengthening data's empowerment of AI development. This article reviews the document and its implications.
On 19 May 2026, at a forum of the China Network Civilization Conference, the National Technical Committee on Information Security (TC260) released the AI Application Ethics and Security Guideline 1.0 (TC260-005), setting ethical and security guidance for the development, provision and use of AI applications. This article reviews its key points.
Media reported on 18 May 2026 that the Shanghai Huangpu District People's Court decided a copyright dispute over AI prompts, holding that a simple list of prompts is not a work and dismissing the claim. The ruling shifts scrutiny from AI output to the starting point of creation and reaffirms that copyright protects expression, not ideas.
On 12 May 2026 the World Digital Education Conference in Hangzhou released the Reference Framework on AI Ethics in Education, giving basic, higher and vocational education a systematic ethical framework and code of conduct, and establishing the principle that the human remains the subject. This article reviews the framework and its implications.
In May 2026 provincial cyberspace administrations continued to publish filing and registration notices under the Interim Measures for Generative AI Services, with Guangdong, Jiangsu and Zhejiang releasing new batches including vertical video and spatial-intelligence models. This article reviews the difference between filing and registration and the practical points.
On 8 May 2026 the Cyberspace Administration of China, the National Development and Reform Commission and the Ministry of Industry and Information Technology jointly issued the Opinion on the Standardised Application and Innovative Development of AI Agents, implementing the State Council's 'AI Plus' deployment. It calls for strengthening data security, personal-information protection, permission management and behaviour control for AI agents, and sets out nineteen typical application scenarios and a tiered governance framework. This article reviews the document's key points and what they require of companies.
On 7 May 2026 the European Parliament and EU member-state representatives reached political agreement on amending the AI Act (the 'AI Omnibus'), agreeing to ban AI systems from generating deepfake pornography, set a clear implementation timetable for high-risk AI rules and simplify innovation-facing rules. This article analyses the parallel 'simplify and draw red lines' approach and what it means for companies going global.
In May 2026 a judgment in the copyright infringement case brought by the Huangpu District People's Procuratorate of Guangzhou, Guangdong, against Qin and Sha became effective. Without the copyright owner's licence, the two defendants copied and recorded more than 1,700 AI-generated short dramas and sold and distributed them, constituting the crime of copyright infringement; they were sentenced to eight months' imprisonment, suspended for one year and two months, and fined. This article reviews the ruling and the shift from civil enforcement to criminal protection for AIGC works.
On 30 April 2026 the Cyberspace Administration of China issued a notice launching a four-month 'Qinglang — Rectifying AI Application Disorder' special action nationwide, reported in the media from 1 May. The action focuses on fourteen problem categories and proceeds in two phases, the first being the 'Qinglang — Typical Violations in AI Application Services' campaign. This article reviews the governance framework and what the categories mean for companies' self-assessment.
In April 2026 MIIT and the National Data Administration jointly issued a notice on implementing the 2026 'Model-Data Resonance' action, setting seven priority tasks to promote coordinated and mutually reinforcing development of AI models and data resources. This article reviews the key points and what they require of AI companies' data compliance and dataset building.
On 29 April 2026 the Ninth Digital China Summit opened in Fuzhou under the theme 'Accelerating Digital-Intelligence Technology Innovation, Deepening Digital China Construction', with more than 50 dialogue sessions. The AI trend on display is a shift from 'talking' to 'doing' — the era of agents and embodied intelligence. This article reviews the industry and governance signals.
On 28 April 2026 the Cyberspace Administration of China disclosed that CapCut, Maoxiang and Jimeng AI had failed to effectively implement synthetic-content labelling requirements and were penalised by cyberspace authorities. This article reviews the regulatory basis, common violations and self-inspection points.
On 27 April 2026 the Foreign Investment Security Review Office (NDRC) made a prohibition decision on a foreign acquisition of the Manus project, requiring the parties to unwind the transaction. This article reviews how foreign-investment security review applies in the AI sector, and what AI companies must assess before cross-border deals and going-global planning.
On 23 April 2026 details of China's first AI model infringement case were disclosed. The court held that the defendant's special-effect model, formed through data training and tuning, constituted a substantial substitute for the plaintiff's model and caused substantial harm to its competitive interests, amounting to unfair competition, with damages of RMB 1.6 million upheld on appeal. This article reviews the reasoning and competition-compliance points.
On 22 April 2026 the Beijing High People's Court released its 2025 Typical Cases on Judicial Protection of New Quality Productive Forces, including a case on the crime of copyright infringement using AI generative models. The court made clear that AI is not a shield against infringement: using AI to copy and distribute others' original artwork for profit meets the elements of the crime. This article reviews the ruling and the compliance path.
On 20 April 2026, at a seminar hosted by the Guangdong High People's Court, 24 AI-sector organisations jointly released a Consensus on Strengthening Collaborative Innovation to Promote the Prosperity of AI Open-Source Ecosystems. This article reviews the judicial background and what it means for companies upstream and downstream of open-source models.
On the evening of 20 April 2026 Moonshot AI released and open-sourced the Kimi K2.6 model, further improving long-horizon coding and agentic execution, with the model going live on third-party inference platforms the next day. This article reviews the compliance points in commercial use of open-source models — licences, downstream filing and ownership of results.
On 13–14 April 2026 the 2026 World Internet Conference (WIC) Asia-Pacific Summit was held in Hong Kong under the theme 'Intelligence Leading Change, New Quality Future — AI-Driven High-Quality Economic Development', with a ministerial meeting, an AI safety governance forum and think-tank outputs. This article reviews the governance signals and what they mean for companies going global.
On 10 April 2026 the Safety Governance Committee of the Artificial Intelligence Industry Alliance (AIIA) held its first 2026 meeting in Wuhan, bringing together universities and leading model and security companies. This article reviews the development of industry-side safety governance coordination and what it means for companies' compliance work.
On 3 April 2026 the National Data Administration opened public consultation on the Guide on Data Property-Rights Registration (Trial) (draft for comment), covering registration-body management, registration procedures and the use of registration certificates, in order to build a unified national data property-rights registration system. This article reviews the key points and what they mean for AI companies' data-asset management.
On 2 April 2026 the National Medical Products Administration issued the Implementation Opinions on 'Artificial Intelligence Plus Drug Regulation' (Guo Yao Jian Zong [2026] No. 6), aiming to build a drug-regulation system integrated with AI by 2030 and setting out seven priority directions for regulatory digitalisation and five foundational tasks. This article reviews the key points and what they mean for pharmaceutical companies and AI service providers.
On 2 April 2026 ten ministries including MIIT jointly issued the Measures on AI Science-and-Technology Ethics Review and Services (Trial) (Gong Xin Bu Lian Ke [2026] No. 75), setting out the scope, responsible bodies, working procedures and supervision of AI ethics review, and distinguishing general, simplified, expert re-review and emergency procedures. This article reviews the key points and their practical implications.
On 1 March 2026 Vietnam's Law on Artificial Intelligence (No. 134/2025/QH15) took effect, replacing the AI provisions of the Law on the Digital Technology Industry and establishing a risk-tiered regulatory framework. This article reviews the law's scope and the compliance points for companies expanding into Southeast Asia.
On 20 March 2026 the White House published a national AI legislative framework setting out recommendations on child safety, intellectual property, preventing censorship, innovation and an AI-ready workforce, aiming to lay the groundwork for a federal standard in Congress. This article reviews the framework and what it means for companies operating in both China and the US.
In March 2026 the European Parliament adopted its position on the AI Act part of the Commission's digital omnibus package, supporting a postponement of certain high-risk AI rules and a ban on 'nudifier' apps. This article reviews the background and what it means for companies going global.
A defamation case in which the defendant argued that 'the video copy was written by AI' has drawn attention. The court held that users of generative AI services bear a duty to verify information when producing and publishing AI-generated content; publishing without citing sources or checking accuracy amounted to infringement. This article reviews the reasoning and the compliance points for content publishing.
On 20 March 2026 the Beijing Internet Court reported a portrait-rights case arising from an AI face-swap short drama that 'resembled' a famous actor. The court held that the production company had infringed the actor's portrait rights by using deep-synthesis technology, and that the broadcasting company, having failed to exercise reasonable review, was liable too. This article reviews the ruling and the compliance points for drama producers and platforms.
On 17 March 2026 the Cyberspace Administration of China announced that, as of 28 February 2026, 796 generative AI services had completed filing and 481 generative AI applications or functions had completed registration, with 48 new filings in January and February. This article reviews the data and the practical points for applicants.
On 12 March 2026 the Cyberspace Administration of China published the sixteenth batch of domestic deep-synthesis service algorithm filings, covering 779 algorithms, including the Doubao enterprise voice-podcast model algorithm, the Qwen image-recognition-generation algorithm and the Turing Zhixin text-generation algorithm. This article reviews the filing rhythm and self-inspection points.
On 12 March 2026 Justice Minister He Rong said at the third 'Ministers' Corridor' session of the 14th NPC's fourth session that AI was the 'high-frequency word' of this year's Two Sessions, and that China will accelerate research on legislation for AI, the low-altitude economy and other fields. This article reviews the background and what it means for companies' compliance expectations.
In March 2026 the Supreme People's Court work report stated that courts should 'properly hear AI-related cases and accurately grasp the room for tolerance for technological innovation', citing a case in which a generative AI service erred but the developer had exercised due care and no rights were actually harmed, so no infringement was found. This article reviews the reasoning and companies' due-care checklist.
On 5 March 2026 the European Commission published the second draft of its Code of Practice on Marking and Labelling of AI-Generated Content, a voluntary guide to help providers and deployers meet the AI Act's Article 50 transparency obligations, promoting open standards and an EU labelling icon. This article reviews the draft and what it means for companies going global.
On 5 March 2026 the Government Work Report first proposed 'building a new form of intelligent economy', from deepening the 'AI Plus' initiative and scaling commercial AI applications in key industries, to cultivating AI-native business models, strengthening the data-element system and improving AI governance. This article reviews three policy signals and what they mean for companies.
On 2 March 2026 the US Supreme Court denied certiorari in Thaler v. Perlmutter (No. 25-449), leaving in place the rule that works generated entirely by AI without human creative contribution are not entitled to copyright protection. This article reviews the case and what it means for AI-generated content and companies going global.
A batch of national standards took effect in China on 1 March 2026, including the two recommended standards GB/T 46069.1-2025 (AI operator interfaces, basic mathematics) and GB/T 46069.2-2025 (AI operator interfaces, neural networks). This article reviews the standards and what they mean for AI hardware-software ecosystems and compliance work.
On 23 February 2026 multiple data protection authorities worldwide jointly issued a statement on artificial intelligence-generated images and privacy protection, responding to the risk that AI image and video generation systems produce identifiable personal content without the individual's knowledge or consent. This article reviews the statement and what it means for image-generation companies.
From 16 to 21 February 2026 the India-AI Impact Summit 2026 was held in New Delhi, the first global AI summit hosted by a Global South country. This article reviews the summit's agenda orientation and what it means for companies going global in reading the global AI governance landscape.
In February 2026 the Cyberspace Administration of China carried out a campaign against accounts that publish AI-generated synthetic information without adding an AI label and use false content to deceive and mislead the public. Some 13,421 accounts were disposed of in accordance with law and agreement, over 543,000 pieces of unlawful information were cleared, and typical cases were reported. This article reviews the enforcement points and directions for corporate self-inspection.
On 7 February 2026 the National Data Administration, the Ministry of Industry and Information Technology, the Ministry of Public Security and the China Securities Regulatory Commission issued the Opinions on Cultivating Data-Circulation Service Institutions and Accelerating the Market-Based Valuation of Data Elements, defining three types of institutions for the first time. This article reviews the key points and what they mean for AI companies' data-source compliance.
On 3 February 2026 the 2026 No. 1 Central Document was released. It calls for developing new agricultural productive forces suited to local conditions, combining artificial intelligence with agricultural development, and expanding application scenarios such as drones, the internet of things and robotics. This article reviews the policy points and what they mean for agricultural AI companies.
On 1 February 2026 the Measures for the Supervision and Administration of Live-Streaming E-Commerce took effect. The measures provide that where live-room operators use figures or videos generated by artificial intelligence and other technologies to conduct live-streaming e-commerce, they must comply with relevant laws and regulations, bringing digital-human streamers and other generated content under regulation. This article reviews the key points and compliance implications for practitioners.
On 31 January 2026 the Cybercrime Prevention Law (Draft for Comment) was released for public comment through the cybersecurity bureau of the Ministry of Public Security. As foundational legislation in the cyber field, its interface with artificial intelligence and data security provisions deserves attention. This article reviews the legislative background and its compliance implications for companies.
On 27 January 2026 the Beijing AI company Moonshot AI released and open-sourced its new Kimi K2.5 model, built on a trillion-parameter mixture-of-experts architecture and natively trained on 15 trillion mixed vision-and-text tokens. This article reviews the compliance points of open-source models from the angles of licence, downstream filing and ownership of results.
On 23-24 January 2026 the 2026 China-ASEAN AI Capacity-Building Workshop and High-Level Seminar on Frontier AI Technology and Governance was held in Kuala Lumpur, Malaysia. With the theme 'people-centred, AI for good', the meeting offered a window on regional AI governance cooperation. This article reviews its signals and what they mean for companies going global.
In January 2026 the National Data Administration issued the Opinions on Strengthening Data Technology Innovation, defining the concept of 'data technology' for the first time and setting out a systematic strategic path and implementation measures. This article reviews the concept and what it means for AI companies' data capability building.
On 18-19 January 2026 the Central Political and Legal Work Conference was held in Beijing. The conference called for cracking down on the use of artificial intelligence to generate and publish disinformation, strengthening research on new technologies, and building digital platforms for political and legal work. This article reviews the AI-related language and what it means for companies.
The Hangzhou Internet Court has concluded China's first network tort dispute arising from a generative AI 'hallucination'. The court held that AI has no legal personality, that inaccurate AI output is governed by the fault-based principle of ordinary tort liability, and that the claim failed for lack of causation and fault. This article reviews the reasoning and key points for companies' risk prevention.
On 14 January 2026 the US Food and Drug Administration and the European Medicines Agency jointly published ten guiding principles on good AI practice in drug development, providing a common framework for using artificial intelligence across the drug-development lifecycle. This article reviews the principles and what they mean for pharmaceutical AI companies.
In January 2026 the Cyberspace Administration of China published 2025 generative AI service filing information: as of 31 December 2025, 748 services had completed filing and 435 applications or functions had completed registration, with 446 new filings during 2025. This article reviews the three trends behind the data.
In January 2026 eight Chinese ministries, led by the Ministry of Industry and Information Technology, jointly issued the Implementation Opinion on the 'AI Plus Manufacturing' Special Action, setting targets to 2027: reliable supply of key technologies, 100 high-quality industrial datasets and 500 typical application scenarios. This article reviews the policy and the governance duties it implies.
On 7 January 2026 the Cyberspace Administration of China published the fifteenth batch of domestic deep-synthesis service algorithm filings. Algorithm filing has become routine and batch-based; companies need to treat it as ongoing state management rather than a one-off action. This article reviews the rhythm and self-inspection points.
On 5 January 2026 Nvidia announced new open models, frameworks and AI infrastructure for physical AI at CES 2026, alongside robots built by global partners. As physical AI moves from the lab toward volume production, three new questions follow: data collection, model licensing and product liability.
On 1 January 2026 the Measures for the Certification of Outbound Personal Information, jointly issued by the Cyberspace Administration of China and the State Administration for Market Regulation, took effect. The measures provide a third compliance route for outbound personal information, alongside security assessment and standard contracts — directly relevant to AI companies that call overseas models or operate across borders.
On 1 January 2026 China's revised Cybersecurity Law took effect — the first amendment since the law was enacted in 2016. A new Article 20 makes dedicated provision for the safety and development of artificial intelligence, lifting AI governance from departmental rules to the level of a foundational statute. This article reviews the changes and what they mean for AI companies.
On 26 December 2025 China's Ministry of State Security issued a public safety notice on AI tools, flagging the risks of misused deepfake technology and disclosing a case in which an employee's improper use of an open-source AI tool led to internal sensitive material being illegally downloaded by a foreign IP address. This article reviews the notice and the data-security arrangements companies should make.
On 18 December 2025 the State Administration for Market Regulation and the Cyberspace Administration of China issued Order No. 116, the Measures for the Supervision and Administration of Online Trading Platform Rules, effective 1 February 2026. The Measures impose procedural requirements on the publication of platform rules, public consultation, consolidation of comments and enforcement. This article reviews the key points for AI service and content platform companies.
On 16 and 17 December 2025 the UN General Assembly held the high-level meeting of the overall review of the implementation of the outcomes of the World Summit on the Information Society (WSIS+20) in New York and adopted an outcome document reaffirming a people-centred, human-rights-based digital future, calling for faster action to close digital divides and highlighting trustworthy governance of data and AI. This article reviews the outcome and what it means for AI companies.
The Central Economic Work Conference was held in Beijing on 10 and 11 December 2025. In setting out next year's priority tasks, it called for deepening and expanding 'AI plus' and improving AI governance, echoing the 'fully implement the AI plus action' in the 15th Five-Year Plan proposals. This article reviews the policy meaning and the signals for AI companies.
On 11 December 2025 the US President signed Executive Order 14365, Ensuring a National Policy Framework for Artificial Intelligence, aiming at a minimally burdensome national AI policy framework and directing the Attorney General to establish an AI Litigation Task Force within 30 days to challenge state AI laws that are unconstitutional, preempted or otherwise unlawful. This article reviews the order and what it means for companies operating in both China and the US.
On 10 December 2025 Korea's Prime Minister chaired a national policy coordination meeting that released measures against false and exaggerated advertising that uses AI to disrupt market order, requiring labelling of AI-generated material and introducing punitive damages of up to five times losses for knowingly spreading false or fabricated information, alongside 24-hour review and emergency blocking. This article reviews the key points for companies going global.
On 8 December 2025 the Brazilian federal government submitted Bill No. 6.237/2025 in an extra edition of the Official Gazette, proposing a National System for the Development, Regulation and Governance of Artificial Intelligence, co-signed by the Ministry of Management and Innovation, the Ministry of Justice and Public Security and the Social Communication Secretariat. The bill advances alongside the Senate's existing AI legislative proposal. This article reviews its orientation for companies going global.
On 6 December 2025 the Cyberspace Administration of China opened public consultation on the draft Measures for the Administration of Network Data Security Risk Assessment, with comments due by 5 January 2026. Drafted under the Data Security Law and the Network Data Security Management Regulations, the draft would establish an annual risk assessment duty for important data processors and a mechanism for submitting and verifying assessment reports. This article reviews the key points for AI companies.
In December 2025 the Shanghai Municipal People's Government General Office issued the Implementation Plan for Strengthening the Open Source System in Shanghai (Hu Fu Ban [2025] No. 33), targeting an initial open source system by 2027 with one to two internationally influential open source communities, 100 open-source commercialisation companies and more than 200 quality open source projects. This article reviews the key points for AI open source companies.
On 2 December 2025 the US District Court for the Southern District of New York, in The New York Times v OpenAI, denied OpenAI's motion for reconsideration and held that a sample of 20 million user chat logs is relevant in its entirety and must fall within discovery. The order pushes AI providers' user conversation data into the litigation spotlight. This article reviews the reasoning and what it means for data retention.
On 19 November 2025 the European Commission proposed a Digital Omnibus on AI, amending the AI Act and related civil aviation rules to simplify implementation for high-risk AI systems, expand testing opportunities and reduce compliance burdens, while postponing the application dates for high-risk systems. This article reviews the proposal and what it means for companies going global.
China's National Intellectual Property Administration issued a decision amending the Patent Examination Guidelines (Order No. 84), retitling section 6 of Chapter 9, Part II to cover invention applications involving algorithm features or business-rule and method features such as AI and big data, refining the examination baseline and adding an exclusion for applications contrary to law or public interest. This article reviews the changes and what they mean for AI patent strategy.
On 12 November 2025 Korea's Ministry of Science and ICT pre-announced the enforcement decree of the Framework Act on AI Development and the Establishment of a Foundation of Trust (Notice No. 2025-970), with a consultation period to 22 December, focusing on labelling of generative AI outputs and duty allocation among responsible parties. The Act took effect on 22 January 2026. This article reviews the key points for companies going global.
On 11 November 2025 the Cyberspace Administration of China published generative AI service filing information. As of 1 November 2025, 611 generative AI services had completed filing and 306 generative AI applications or functions had completed registration. The parallel growth of filings and registrations shows the compliance focus shifting from the model layer to the application layer. This article reviews the data and the practical points.
On 11 November 2025 the Munich Regional Court I ruled in GEMA v OpenAI that the model's 'memorisation' of lyrics and the chatbot's output of lyrics constituted copyright infringement. The ruling is an important precedent in European AI copyright law. This article reviews the reasoning and the implications for training-data compliance.
From 6 to 9 November 2025 the World Internet Conference Wuzhen Summit was held, releasing the Global AI Standards Development Report and Building a Global AI Safety and Governance System for the Common Good of Humanity. Both point to worsening fragmentation in global AI governance and recommend focusing on standards, safety, inclusion and trust. This article reviews the governance implications.
On 6 November 2025 the Cyberspace Administration of China published the 14th batch of domestic deep synthesis service algorithm filings. The steady release of batches shows that deep synthesis algorithm filing has moved from a one-off clean-up to routine administration. This article reviews the filing mechanism and the self-inspection points for companies.
On 5 November 2025 India's Ministry of Electronics and Information Technology released the India AI Governance Guidelines, setting out a governance framework around safe, trusted, inclusive and accountable AI, with principles on transparency, fairness, accountability, risk assessment and the copyrightability of generative AI outputs. This article reviews the guidelines and what they mean for companies going global.
On 1 November 2025 the national standard Cybersecurity Technology — Basic Security Requirements for Generative Artificial Intelligence Services (GB/T 45654—2025) took effect, converting the principle-based requirements of the Interim Measures for Generative AI Services into testable technical clauses. This article reviews the standard's positioning and the practical points for companies.
In October 2025 several Chinese ministries jointly issued an implementation opinion on promoting and regulating 'AI plus healthcare' applications. It targets a group of vertical clinical models for specific diseases by 2027, near-universal AI-assisted primary care and widespread AI-assisted medical imaging diagnosis at secondary hospitals and above by 2030, and an AI application standards system. The compliance boundary for medical AI is becoming clearer.
From 20 to 23 October 2025 the Fourth Plenary Session of the 20th Central Committee adopted the Recommendations for the 15th Five-Year Plan. The Recommendations treat AI as a focus of digital and intelligent technology innovation and require strengthening AI governance by improving laws, regulations, policy systems, application norms and ethical guidelines — providing the top-level basis for AI legislation.
From 16 to 18 October 2025 the 2025 World Intelligent Connected Vehicle Conference was held. It signalled faster development of standards for combined driving assistance and autonomous driving, stronger production consistency supervision for vehicles with such systems, and tighter rules for over-the-air software updates. The compliance focus in AI-plus-automotive is shifting from features to responsibility.
On 14 October 2025 the European Data Protection Board announced that its 2026 coordinated enforcement framework will focus on GDPR transparency and information duties. National data protection authorities will jointly examine whether controllers adequately inform individuals about their data processing. Automated decision-making and disclosure in AI services now enter unified enforcement.
In October 2025 China's Ministry of Industry and Information Technology launched a city-area 'millisecond computing' initiative, aiming by 2027 for a city-area compute network with millisecond interconnection between compute centres and millisecond reachability for application terminals. Compute policy is shifting from scale to latency.
In October 2025 China's cyberspace regulator and development planner jointly issued guidance on deploying AI large models in government affairs. It positions government large models as assistive, requires measures against model hallucination, sets a strict confidentiality requirement against inputting classified information into non-classified models, and warns against 'digital formalism'.
In October 2025 the Financial Stability Board published a report on monitoring AI adoption and related risks in the financial sector, advising regulators on how to monitor AI adoption, risk management and vulnerabilities at financial institutions. Though not binding, the report transmits into practical constraint through supervisory expectations and peer practice.
In October 2025 the UK Upper Tribunal ruled in the Clearview AI case, upholding three of the ICO's grounds of appeal and confirming that Clearview's processing was linked to monitoring UK residents and that it could not escape UK data protection law by serving only foreign law enforcement. The ruling reinforces the extraterritorial reach of the 'behaviour monitoring' test.
In October 2025 the G7 Cyber Expert Group issued a statement on AI and cybersecurity, focusing on risks from generative and agentic AI and recommending risk assessment, supply chain management and AI software bills of materials for the financial sector. The statement formally connects AI safety with cybersecurity governance.
In October 2025 China's market regulator and standards authority released a batch of AI national standards covering AI risk management capability assessment, federated learning technical specifications, deep learning compiler interfaces, computing centre capability assessment, and general technical requirements for customer-service virtual digital humans. The standards system is extending from 'may we use it' to 'how to use it safely'.
On 29 September 2025 California's governor signed SB 53, the Transparency in Frontier Artificial Intelligence Act. It requires large frontier AI developers to publish safety frameworks, report critical safety incidents, protect risk disclosure, and authorises civil penalties. With federal rules still unsettled, state law is becoming the practical driver of US AI governance.
In September 2025, one year after trial operation, the national computing network service platform released results including a 31-province identifier system and 52 data express routes. As compute and data interconnectivity improves, the compliance questions around compute identifiers, cross-region scheduling and data flows come to the fore.
On 23 September 2025 the 2025 AI Industry and New Industrialisation Conference opened in Beijing, with more than 500 delegates from industry, academia, research, application and finance, and several landmark results released. The signal is that AI policy is shifting from building general capability towards deep integration with the real economy, with industrial settings as the next battleground.
In September 2025 a court in Shangyu, Zhejiang, delivered first-instance judgment in a case where defendants used AI face-swap technology to bypass identity verification and buy Maotai liquor. They were convicted of illegally controlling computer information systems and of infringing citizens' personal information. The case shows that misuse of face-swap technology often crosses two criminal lines at once.
In September 2025 China's National Technical Committee 260 on Cybersecurity released a technical document, Security Specification for Government Large Model Applications, setting security requirements and test methods across model selection, deployment, operation and decommissioning. It gives AI vendors serving the public sector a clear benchmark.
On 11 September 2025 the Cyberspace Administration of China published the 13th batch of domestic deep synthesis service algorithm filings. The steady cadence shows deep synthesis filing has moved from early concentrated filings into routine batch management, and companies should adjust their expectations on filing, change and deregistration duties accordingly.
In September 2025 the Supreme People's Court released a set of typical unfair competition cases, including one holding that an AI model built through data training and parameter tuning is a protected competitive interest, and that directly using another party's model structure and parameters constitutes unfair competition. The ruling opens a path to protecting AI model assets outside copyright.
On 1 September 2025 the Measures for Labelling AI-Generated Synthetic Content, jointly issued by four Chinese regulators, took effect, together with the supporting mandatory national standard. Labelling moves from an industry expectation to a legally binding duty, and the responsibilities of service providers and content platforms are now clearly divided. This article reviews the key points and what companies should do.
On 1 September 2025 Japan's Act on the Promotion of Research, Development and Utilisation of AI-Related Technologies came fully into force. The statute puts innovation first, sets out three principles — human-centric, risk minimisation and innovation promotion — and establishes an AI Strategy Headquarters. Its promotion-led approach contrasts with the EU's regulation-led model.
On 27 August 2025 the Ministry of Transport outlined the '1+N+X' overall technical architecture for an integrated transportation large model at its regular press conference: '1' is a shared general technical foundation enabling open compatibility with multiple base models and heterogeneous computing power; 'N' is a set of domain models trained on that foundation; and 'X' is agents serving specific business scenarios. The MOT also catalogued 860 typical AI application scenarios, with the first agents to be deployed on the National Integrated Transportation Information Platform.
On 26 August 2025 the launch and discussion meeting for the national standardisation guiding technical document 'General Rules for Artificial Intelligence-Assisted Contract Drafting and Review Services' was held at the China National Institute of Standardization. The document, plan number 20251506-Z-469, is administered by the National Technical Committee 264 on Service Standardization (SAC/TC 264) and drafted by bodies including the China National Institute of Standardization and Fatianshi. AI contract services have thus entered a standardisation track.
On 26 August 2025 the State Council issued the Opinions on Deepening the Implementation of the 'AI Plus' Action, setting three-phase goals: by 2027 adoption of next-generation smart terminals and AI agents should exceed 70%, by 2030 exceed 90%, and by 2035 China should fully enter a stage of intelligent economy and intelligent society. The Opinions deploy tasks across six priority sectors and call for strengthened coordination of intelligent computing power.
On 20 August 2025 the UK Information Commissioner's Office published an audit report on the use of facial recognition by South Wales Police and Gwent Police, covering both live facial recognition and retrospective matching. It is the ICO's latest move in sustained pressure on biometrics, one of the most sensitive AI scenarios, and offers a compliance reference for public bodies deploying facial recognition.
On 19 August 2025 a new batch of telecoms industry standards in the YD/T 4392 series was released, including 'General Capability Requirements for Artificial Intelligence Development Platforms - Part 2: Security Requirements' and 'Part 4: Large Model Development and Application', effective 1 December 2025. Security and engineering capability requirements for AI development platforms are entering the industry standards system.
On 8 August 2025 the Hangzhou Intermediate People's Court ruled in Zhejiang's first unfair-competition case involving the commercial application of AI. It held that an AI writing tool offering bulk copy generation for social e-commerce 'seed planting' scenarios constituted unfair competition, ordering two internet companies to pay RMB 100,000 jointly. The ruling makes clear that technology neutrality does not equal application neutrality.
On 2 August 2025 the EU AI Act's obligations for general-purpose AI models took effect. All GPAI providers must publish a training-data summary and adopt a copyright compliance policy; providers established outside the EU must appoint an EU representative; and the most advanced models posing systemic risk face additional evaluation, adversarial testing and serious-incident reporting duties.
From 1 August 2025 the Cyberspace Administration of China ran a 20-day 'Qinglang - Chengdu World Games online environment' campaign. It lists six categories of prominent problems and expressly names the use of AI to create and publish fake event videos as a target, making major sporting events a high-pressure test of AI content governance.
On 31 July 2025 the State Council executive meeting reviewed and approved the Opinion on Deepening the Implementation of the AI+ Action. The meeting called for vigorously advancing large-scale, commercialised AI applications, drawing on China's complete industrial system, large market and rich application scenarios, promoting faster adoption and deep integration of AI across the economy and society, and requiring government departments and state-owned enterprises to strengthen their demonstration role.
On 28 July 2025 Shanghai issued Several Measures to Further Expand AI Applications, offering RMB 1 billion in special subsidy vouchers: RMB 600 million in compute vouchers, RMB 300 million in model vouchers and RMB 100 million in corpus vouchers, alongside talent-service measures. The vouchers bring the cost of the three core AI factors directly into the scope of fiscal support, a textbook example of factor-subsidy industrial policy.
From 26 to 28 July 2025 the World AI Conference and the High-Level Meeting on Global AI Governance were held in Shanghai. The Chinese government proposed establishing a World AI Cooperation Organisation, with Shanghai as the initial candidate for its headquarters, and the meeting released a Global AI Governance Action Plan setting out 13 concrete actions. It is a moment when AI governance moves from debating rules to building mechanisms.
On 24 July 2025 the General Office of the Ministry of Human Resources and Social Security issued a notice on broadly delivering AI literacy continuing education for professional and technical personnel. The learning content covers technical foundations such as algorithms, compute and data, and expressly includes AI ways of thinking, ethics and law. AI literacy is turning from a specialist skill into a general capability.
On 24 July 2025 the Cyberspace Administration of China launched the two-month Qinglang campaign against false information posted by self-media. It expressly targets the use of AI generation and synthesis technology to impersonate others and fabricate false information, and requires that source attribution and AI-generated content labelling be improved — information that should be labelled but is not may not enter the algorithm recommendation pool.
On 23 July 2025 the White House released Winning the AI Race: America's AI Action Plan, listing more than 90 federal policy recommendations, and signed several executive orders the same day. The directions include value-based federal AI procurement, promotion of US AI technology-stack exports, and faster permitting for data-centre infrastructure. The centre of gravity of US AI policy has shifted from risk constraint to competitiveness first.
On 21 July 2025 the Shenzhen Municipal Government Services and Data Management Bureau issued the Operating Rules for the Special Fund for Open Government AI Scenario Applications, effective the same day. The fund provides up to RMB 50 million a year and supports, in graded tiers, companies that build AI applications on open government scenarios. This is the open-scenario-plus-fiscal-funding tool of industrial policy applied concretely to AI.
In July 2025 the Cyberspace Administration of China launched the two-month Qinglang 2025 Summer Campaign on Minors' Online Environment. It targets four categories of problems and explicitly brings the improper use of AI features for minors, and AI-induced overuse, into scope. Minors' mode, content safety on children's smart devices and AI companion applications are examined side by side for the first time.
On 14 July 2025 the Cyberspace Administration of China published the twelfth batch of domestic deep synthesis service algorithm filings. The announcement reiterates that providers of deep synthesis services with public-opinion attributes or social-mobilisation capacity must complete filing, change and cancellation procedures under the algorithm recommendation rules, and that technical supporters must do the same by reference. Filing has moved from a one-off registration to routine, batch-based administration.
On 10 July 2025 the European Commission published the General-Purpose AI Code of Practice. A voluntary instrument, it lets GPAI model providers show that they meet the relevant AI Act obligations. Built around transparency, copyright, and safety and security, it became the most important compliance interface before the GPAI obligations took effect on 2 August.
In July 2025 the General Office of the Ministry of Education issued a notice on implementing the Digital-Empowerment Teacher Development Action, making the integrated and innovative use of digital and AI technologies the driver of teacher development. The document implements the education digitalisation strategy and the AI+ agenda in education, and puts teachers' AI capability, content-labelling awareness and ethical norms on the agenda.
On 30 June 2025 the national standard GB/T 45907-2025 on assessing the maturity of AI service capability was issued and took effect. It sets out a capability framework, grades maturity across five levels from initial to outstanding, and supplies scoring and assessment methods for evaluating a provider's AI service capability as a whole or domain by domain.
On 25 June 2025 NIST published NIST AI 700-1, the results of its 2024 GenAI pilot study on text generation and discrimination. AI-written summaries are moving closer to human writing, yet detection models remain relatively effective; performance varies sharply between systems, with generators that fool most discriminators and discriminators that catch nearly all generators.
On 23 June 2025 the National Technical Committee 260 on Cybersecurity opened public consultation on six cybersecurity standard practice guides, including one on implicit labelling of text files through metadata. They are the implementation detail for the Measures for Labelling AI-Generated Synthetic Content and the mandatory standard GB 45438-2025.
On 20 June 2025 the Cyberspace Administration of China published phase one results of the Qinglang campaign against the misuse of AI technology: more than 3,500 non-compliant AI products handled, over 960,000 items of unlawful information cleared and more than 3,700 accounts dealt with. Phase two will focus on seven categories of problem, including AI-driven rumour and vulgar content.
On 17 June 2025 G7 leaders issued the AI for Prosperity statement, stressing the role of safe, trusted AI in growth and jobs and pushing for a G7 AI application roadmap and a G7 AI network (GAIN) to accelerate adoption in the public sector and among small and medium-sized enterprises.
On 13 June 2025 the Hangzhou municipal government issued its 2025 implementation plan for building an AI innovation highland. By the end of 2025 the city aims to bring more than 50 EFLOPS of intelligent computing to market and to cultivate two foundation models and at least twenty-five vertical application models.
On 5 June 2025 the UK Information Commissioner's Office published a new AI and biometrics strategy covering foundation model training, generative AI, automated decision-making and biometrics. It argues for stronger front-end supervision so that organisations comply with data protection law while developing and using AI.
In June 2025 MIIT studied how to advance the AI industry and enable new industrialisation, and at a subsequent meeting on industrial-digital integration required an AI plus manufacturing initiative, faster intelligent upgrading of key industries, progress on deep synthesis detection technology and new rules on AI technology ethics management.
On 30 May 2025 the Cyberspace Administration of China published a policy Q&A on outbound data security management. It addresses how important data is identified and declared and how the outbound assessment works, and confirms that a processor told or publicly notified that its data is important must apply for a security assessment within two months.
On 28 May 2025 the Cyberspace Administration of China issued an announcement on filing for the application of face recognition technology. From 1 June 2025, anyone storing face information on 100,000 people must file within thirty working days; those already over the threshold by 1 June had to complete filing by 14 July.
In May 2025 the Beijing Municipal Bureau of Economy and Information Technology issued its 2025 action plan for AI-enabled industrial modernisation. It offers up to RMB 30 million in compute cost support for industry models at a leading domestic or international level, and backs the building of general-purpose agents with industry-wide reach.
On 22 May 2025 the Cyberspace Administration of China launched a two-month Qinglang campaign to improve the online business environment by tackling so-called black mouths against companies: profit-driven false corporate information, malicious smearing and extortion. AI-generated content has lowered the cost of rumour, making corporate information governance part of the business environment.
In May 2025 MIIT issued its Action Plan for Compute Interconnection. By 2026 it aims for a fairly complete system of standards, identifiers and rules for compute interconnection, national, regional and industry interconnection platforms, and a unified pool of public compute identifiers enabling cross-operator, cross-architecture and cross-region scheduling of supply and demand.
On 19 May 2025 the Cyberspace Administration of China published the eleventh batch of domestic deep synthesis service algorithm filings. The notice again confirms that providers of deep synthesis services with public opinion attributes or social mobilisation capacity must file, with technology providers following by reference. Filing has become a periodic ledger.
On 19 May 2025 the United States signed the TAKE IT DOWN Act. It brings non-consensual distribution of intimate images, including AI-generated synthetic content, within its scope, and requires platforms to remove the material within 48 hours of receiving valid notice. Deepfake governance moves from platform self-regulation to a statutory duty.
On 12 May 2025 the basic education teaching guidance committee of the Ministry of Education issued a 2025 AI literacy guide for primary and secondary schools and a separate guide on students' use of generative AI. Together they set out, for the first time, stage-by-stage rules for using generative AI in basic education.
On 9 May 2025 the US Copyright Office published the pre-publication version of part three of its Copyright and Artificial Intelligence report, on generative AI training. It argues that whether training on copyrighted material is fair use must be judged case by case, weighing purpose and market effect, rather than by a blanket rule.
On 1 May 2025 the Hangzhou ordinance promoting the development of the embodied AI robot industry took effect. It establishes a legal definition of embodied AI robots and introduces coding management and full-life-cycle traceability, an early piece of dedicated local legislation that writes promotion and regulation into the same text.
In April 2025 the Beijing Economic-Technological Development Area issued its 2025 implementation plan for building a city-wide artificial intelligence district, introducing a 'one model, one plan; one industry, one policy' mechanism. By the end of 2025 it aims to open 100 landmark application scenarios, cluster 600 core companies in the industry chain and exceed RMB 80 billion in industry scale. The centre of gravity in local AI policy is moving from cash subsidies to scenario supply.
On 25 April 2025 the Political Bureau of the CPC Central Committee held its twentieth collective study session on strengthening the development and regulation of artificial intelligence. The session stressed self-reliance and an application-oriented approach, called for AI to develop in a healthy and orderly way in a beneficial, safe and fair direction, and pointed to the advantages of the new system for mobilising resources nationwide.
In April 2025 the Ministry of Education and eight other departments jointly issued the Opinion on Accelerating Education Digitalisation (Jiao Ban [2025] No. 3). It is the first time education digitalisation has been issued jointly by nine departments, bringing the use of AI in teaching, management and assessment under a single deployment and writing data security and platform standards into the same document.
On 10 April 2025 US senators reintroduced the Content Origin Protection and Integrity from Edited and Deepfaked Media Act (COPIED Act). The bill would establish federal transparency standards for content provenance, requiring AI-generated content to be marked, authenticated and detectable, and would protect journalists, artists and songwriters.
On 9 April 2025 the European Commission published the AI Continent Action Plan, setting out directions for building large-scale AI compute and data infrastructure, improving data accessibility, driving AI adoption in strategic industries and strengthening skills and talent. It is the EU's move to fill in the supply-side gaps that sit outside the AI Act.
On 9 April 2025 US senators introduced S.1367, the Protecting the Voice and Likeness of Individuals Act, referred to the Senate Judiciary Committee. The bill seeks to give individuals federal-level protection over their voice and likeness, akin to intellectual property, in response to the infringement risks raised by AI voice cloning.
On 8 April 2025 the Beijing Municipal Bureau of Economy and Information Technology issued its 2025 action plan to support information software companies in strengthening AI application service capability, setting out eight measures. Support of up to RMB 30 million applies respectively to MaaS platform compute deployment cost, non-hardware procurement for industry models and the first listing of general-purpose agents.
In March 2025 the State Administration for Market Regulation disclosed progress on AI national standards. Since the start of 2025 it has issued national standards on general requirements for AI large models, evaluation metrics and methods, and service capability maturity assessment, and is accelerating standards for deep learning compilers, high-quality datasets and compute scheduling and coordination.
On 27 March 2025 the US Federal Trade Commission filed suit against Cleo AI, Inc., alleging false advertising about the timing and amount of its cash advance service, hidden fees and a subscription mechanism that was difficult to cancel. The case shows that traditional consumer protection rules still apply to the advertising and subscription design of AI apps.
On 26 March 2025 US representatives introduced H.R. 2385, the CREATE AI Act, aimed at establishing a National AI Research Resource (NAIRR) and lowering the barrier to AI research through open research tools, infrastructure and data. The bill reflects the US policy orientation on supplying basic research resources.
In March 2025 the Shanghai Municipal Commission of Economy and Informatisation issued its 2025-2027 implementation opinion on promoting innovation in the intelligent computing cloud industry. By 2027 it aims for an industry scale exceeding RMB 200 billion, intelligent computing capacity of 200 EFLOPS with an autonomously controllable share above 70 percent.
On 21 March 2025 US representatives introduced H.R. 2246, the FIGHT China Act, focused on US persons' investment in sensitive Chinese sectors including artificial intelligence, and seeking to restrict the flow of capital through prohibition or mandatory notification. The bill reflects the continuing tightening of national security review in the AI field.
In March 2025 the Shenzhen Municipal Bureau of Industry and Information Technology issued operating rules for the AI pioneer city project support programme, covering compute support, foundational data, AI software, AI products and AI services. Ten projects are supported, including model vouchers, industry application demonstration benchmarks and domestic AI ecosystem source innovation centres, on a post-hoc reward basis.
On 12 March 2025 the Cyberspace Administration of China published the tenth batch of domestic deep synthesis service algorithm filings. Citing the Provisions on the Administration of Deep Synthesis of Internet Information Services, the notice restates that providers of deep synthesis services with public opinion attributes or social mobilisation capacity must complete filing, change and cancellation procedures, with technology providers following by reference.
On 7 March 2025 the European Commission adopted Implementing Regulation (EU) 2025/454, setting out the establishment and operation of the scientific panel of independent experts under the EU AI Act. The regulation is an early step in the Act's implementation architecture and, together with the AI Board and the advisory forum, supports the later application of general-purpose model obligations.
On 5 March 2025 the government work report proposed continuing to advance the AI Plus initiative, stressing that digital technology should be better combined with manufacturing and market advantages and that wide application of large models should be supported. It also deployed next-generation intelligent terminals such as intelligent connected new energy vehicles, AI phones and computers and intelligent robots, as well as intelligent manufacturing equipment.
On 3 March 2025 the OECD published Towards a Common Reporting Framework for AI Incidents, offering a shared baseline for global AI incident reporting. The framework aims to make AI accidents and risk events comparable across jurisdictions, so that they can serve risk management and policy making.
On 1 March 2025 the national public data resource registry platform officially went into operation, and the interim measures on the administration of public data resource registration took effect at the same time. The platform adopts a one-standard, two-tier architecture, with the national platform and provincial platforms carrying out registration work in coordination, providing a basis for confirming rights in and circulating public data resources.
On 21 February 2025 the Cyberspace Administration of China published the enforcement priorities for its 2025 Qinglang campaign. Among the eight tasks, 'rectifying the misuse of AI technology' appears as a standalone item for the first time, stressing both technology-side and content-side management, stronger labelling of synthetic content, and action against AI-generated disinformation and paid posting.
Between late January and mid-February 2025, Italy's data protection authority blocked DeepSeek for insufficient answers on data sources and legal basis, South Korea restricted app-store downloads and barred access on government devices, and Australia issued a government-device ban. Generative AI going overseas now faces two parallel lines of review: data compliance and national security.
In February 2025 the Cyberspace Administration of China published measures on personal information protection compliance audits, effective 1 May 2025. Processors handling personal information of more than 10 million people must carry out a compliance audit at least once every two years; where a security incident affects over one million people, regulators may require an audit by a professional institution.
On 13 February 2025 a US representative introduced H.R. 1283, the Protecting Our Children in an AI World Act of 2025, which would prohibit the use of artificial intelligence to produce child sexual abuse material. Together with the TAKE IT DOWN Act, it forms a legislative push on non-consensual intimate imagery and deepfakes.
On 11 February 2025 Singapore's Ministry of Digital Development and Information announced a new round of AI safety initiatives, including a Global AI Assurance Pilot for generative AI applications, a joint testing report with Japan, and an evaluation report from the AI Safety Red Teaming Challenge, reinforcing testing and evaluation as a foundation of AI governance.
The AI Action Summit was held in Paris on 10 and 11 February 2025 and closed with a Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet. More than 60 countries and international organisations, including France, China, India and the EU, signed; the United States and the United Kingdom did not.
On 4 February 2025 Australia's Department of Home Affairs issued PSPF Direction 001-2025 under its protective security policy framework, banning the use of DeepSeek products, applications and network services in government entities and in projects receiving government funding. Generative AI applications are being drawn into routine government-device and supply-chain security review.
On 2 February 2025 the provisions of the EU AI Act (Regulation (EU) 2024/1689) on prohibited practices and AI literacy began to apply. On 4 February the European Commission published guidelines on prohibited practices, explaining harmful manipulation, social scoring and real-time remote biometric identification, with examples.
On 20 January 2025 the US President signed an order rescinding the Biden-era AI executive order EO 14110; on 23 January he signed EO 14179, Removing Barriers to American Leadership in Artificial Intelligence, requiring a review of existing AI regulatory actions and shifting toward a policy framework led by private-sector innovation.
On 23 January 2025 Zhejiang province said at its 2025 economic policy briefing that it would issue an AI Plus action plan for 2025 to 2027, further strengthening support for artificial intelligence. Policy supply at the local level around compute, models and application scenarios continues to intensify.
On 21 January 2025 South Korea enacted the Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. 20676), which enters into force one year after enactment, to build a safe and trustworthy foundation for AI development. The accompanying enforcement decree later detailed obligations such as labelling of generative AI outputs.
On 20 January 2025 the MIIT General Office and the NMPA Department of General Affairs launched the 2025 AI medical device innovation challenge, covering three task categories: intelligent decision-support products, brain-computer hybrid intelligence products, and supporting environments, to deepen the integration of AI with medical devices.
On 19 January 2025 the Cyberspace Administration of China launched a one-month campaign on the Spring Festival online environment, targeting six problem areas and stating that the use of AI tools to fabricate false imagery tied to Spring Festival topics to mislead users falls within scope.
On 15 January 2025 the US Bureau of Industry and Security published the Framework for Artificial Intelligence Diffusion (90 FR 4544), amending export controls on advanced computing integrated circuits and adding controls on certain advanced closed-weight AI model weights. Compliance requirements were originally set to take effect on 15 May 2025.
On 13 January 2025 the National Development and Reform Commission and other departments issued an opinion on promoting the high-quality development of the data labelling industry, calling for key technology work on cross-domain and cross-modal semantic alignment, 4D labelling and large-model labelling, and supporting intelligent tools for multimodal labelling, label review and quality assessment.
On 8 January 2025 the National Development and Reform Commission and the National Data Administration issued trial rules on the authorised operation of public data resources, effective 1 March 2025. Together with the interim measures on registration of public data resources, they build a registration-authorisation-operation chain.
In January 2025 eight Beijing departments, including the municipal science and technology commission, issued trial implementation opinions on strengthening science and technology ethics governance. Anchored in Beijing's role as an international centre for science and technology innovation, they explore a 'Beijing approach' to ethics governance in artificial intelligence and biomedicine.
On 1 January 2025 the Regulations on the Administration of Network Data Security (State Council Order No. 790) took effect, regulating network data processing activities, setting out security duties for network data processors and personal information protection requirements, and providing baseline rules for important data, cross-border transfer and platform responsibility.
On 30 December 2024 the National Data Administration released the first batch of Commonly Used Terms in the Data Field, giving unified definitions for data, data products and services, data processing, data security and data analysis — a shared vocabulary for a market that has grown faster than its definitions.
On 26 December 2024 Japan's AI Strategy Headquarters and the AI Institutional Research Group held a joint meeting to discuss the interim summary of AI-related institutional design, confirming Japan's continued preference for a soft-law, guideline-led approach over hard statutory obligations.
On 24 December 2024 the 79th session of the UN General Assembly adopted a resolution on the military applications of artificial intelligence and their implications for international peace and security — non-binding, but the first time the UN's main deliberative body has treated military AI as a distinct topic.
On 20 December 2024 Italy's data protection authority, the Garante, fined OpenAI EUR 15 million, concluding its investigation into the use of personal data to train ChatGPT — one of the largest GDPR penalties aimed at a generative AI provider, and a decision that turns on legal basis and transparency rather than on the model itself.
On 20 December 2024 the General Office of the Shanghai Municipal People's Government issued the implementation plan for AI 'Model the City' (Hu Fu Ban Fa [2024] No. 27), advancing AI applications across city building and industry and positioning Shanghai as a testbed for large-scale deployment.
On 20 December 2024 the Cyberspace Administration of China published the ninth batch of deep synthesis service algorithm filings, confirming that filing has become the routine entry requirement for providers of synthetic media, voice cloning and face-swap services in the Chinese market.
On 18 December 2024 Shenzhen's industry and information technology bureau issued 18 measures to build the city into an AI pioneer city (Shen Gong Xin Gui [2024] No. 13), combining computing supply, model development and application support into a single industrial-policy package.
The Central Economic Work Conference, held 11-12 December 2024, set priorities for 2025 and explicitly called for an 'AI Plus' (人工智能+) initiative — the signal that AI had moved from a sector topic to a macroeconomic instrument in China's planning.
Beijing Economic-Technological Development Area (Beijing E-Town) revised and issued policies to build an AI-native industry innovation highland (Jing Ji Guan Fa [2024] No. 29), targeting 100 core companies across the AI chain and industry scale above RMB 100 billion by 2026.
On 24 November 2024 four Chinese authorities launched the 'Qinglang · Algorithmic Governance' special campaign, running to 14 February 2025, targeting filter bubbles, manipulated rankings and algorithmic harms to gig workers — the first Qinglang campaign aimed at algorithms themselves rather than at content.
In November 2024 the National Data Administration issued the Trusted Data Space Development Action Plan (2024-2028) (Guo Shu Zi Yuan [2024] No. 119), a framework for moving data across organisational boundaries in a controllable, auditable and compliant way.
The 2024 World Internet Conference Wuzhen Summit was held 19-22 November 2024 under the theme of embracing a people-centred and AI-for-good digital future, with AI governance as one of its central topics.
On 14 November 2024 the European Commission's AI Office published the first draft of the General-Purpose AI Code of Practice, developed through multi-stakeholder consultation and workshops with GPAI model providers, ahead of the AI Act obligations applying to general-purpose AI models.
Six Chinese ministries, including MIIT, the NDRC, the Ministry of Finance, SASAC, SAMR and the National Data Administration, jointly launched the 2024 graded smart-factory cultivation action (Gong Xin Ting Lian Tong Zhuang Han [2024] No. 399), setting out levels for factory intelligence and a cultivation pathway for manufacturers.
On 24 October 2024 the White House issued its first National Security Memorandum on AI (NSM-25), directing the use of AI for national security objectives while establishing governance requirements for frontier AI development and deployment.
On 15 October 2024 Singapore's Cyber Security Agency published the Guidelines on Securing AI Systems and a Companion Guide, proposing a secure-by-design approach that spans the full lifecycle of an AI system rather than only its deployment.
On 10 October 2024 MIIT and the Beijing municipal government held a symposium on AI empowering a new type of industrialisation, deepening ministry-city cooperation on AI and industrial application.
On 3 October 2024 the US Office of Management and Budget issued Memorandum M-24-18, Advancing the Responsible Acquisition of Artificial Intelligence in Government, requiring federal agencies to manage risk, performance and market competition when acquiring AI.