China Case

Regulation

China’s response to the onset of AITs stands out in several ways, but arguably mostly for its regulatory and legislative speed. In the supposed race to set global regulatory standards, China has made an ambitious bid with a flurry of policy enactments over the last decade. China commenced a new era of regulatory activity around AITs beginning with the New Generation AI Development Plan (AIDP) in 2017 [1].  Its purpose, as set out by the State Council, was to develop a strategy that would see China become the world leader in AI by 2030. The Plan identifies AI as key to challenge current and arising issues like economic growth, national security, and technological innovation. It proposes as solutions cultivating internal research and development capacities via building out greater AI infrastructure.

Two years later in 2019, the New Generation AI Governance Expert Committee (established by the Ministry of Science and Technology MOST) published Governance Principles for a New Generation of Artificial Intelligence.[2] This document introduces eight core principles for the development and regulation of 'responsible AI': harmony and friendliness, fairness and justice, inclusivity and sharing, respect for privacy, secure/safe and controllable, shared responsibility, open collaboration, and agile governance.[3]  This served as a foundation for the Ethical Norms for New Generation Artificial Intelligence (ENGAI) published by the same committee in 2021.[4]  ENGAI is responsible for implementing the 2019 governance principle in a 'detailed manner' by providing 'ethical guidance to natural persons, legal persons, and other related institutions engaged in AI-related activities'.[5]  ENGAI is comprised of five sections that total 25 Articles, each of which provide ethical guidance on a variety of issues from research and development to AI management.

An important white paper was published by the Big Data Security Standards Special Working Group of the National Information Security Standardization Technical Committee in 2019.[6]  The authors note that 'AI security standardization is an important component of the development of the AI industry' because 'It plays a fundamental, normative, and leading role in stimulating healthy and benign AI applications and in promoting the orderly and healthy development of the AI industry'.[7] The document is concerned with a range of security issues ranging from protecting critical information infrastructure to managing data security risks in AI development.

A signature policy to have come out of the Cyber Security Administration (CAC) is the Internet Information Service Algorithmic Recommendation Management Provisions, jointly issued with the State Internet Information Office, the Ministry of Industry and Information Technology, the Ministry of Public Security and the State Administration for Market Regulation. A draft proposal was first released on 27 August of 2021, followed by a public comment period until 26 September of that year. A finalised version was published on 4 January 2022, and the law went fully into effect on 1 March 2022. The IISARMP consists of several dozen rules or provisions that regulate the development and use of internet algorithmic recommendation services with wide-reaching implications. The law codifies technical and policy requirements, ethical requirements, and prohibited behaviour for algorithmic providers and operators.

A central theme running through the IISARMP is enhancing the capacity of individuals to interact with algorithms or platforms on their own terms. Considerable emphasis is placed on the 'protection of user rights', which includes appeals to user notification, norms disclosure, procedures for obtaining consent, provisions for opting out of monitoring and surveillance, and control over personal data. These protections constitute a considerable set of tools for individuals to circumscribe and resist the power of algorithms and contest their outputs. This may have major implications for workers as labour is increasingly pushed into digitalised space. Workers confront a future characterised by heightened surveillance and management by smart technologies, and key provisions within the IISARMP could be leveraged by employee representatives to limit and, in some cases eliminate, exploitative and dominating workplace practices. Examples of this include deleting user tags aimed at personal characteristics, demanding explanation of algorithmic practice that influence worker interests, as well as amplified enforcement of existing labour rights.

Another core theme of the IISARMP is a commitment to aligning algorithmic recommendation services to the promotion of the 'common good'. This objective underscores the Chinese government’s intention to implement a social model of governance that directly oversees the absorption of algorithmic and AI technologies and is reflected in the IISARMP’s commitments to increased social supervision, heightened transparency, and the instalment of clearly defined enforcement mechanisms and bodies. Furthermore, the duties imposed on algorithmic recommendation service providers mandate explicitly socially oriented observances like advancing 'the social public interest', respecting 'national security', not 'upsetting the economic or social order', and preventing 'addiction or excessive consumption' among other imperative related to the general welfare. Again, this policy goal may be utilised to promote worker well-being because improved working conditions is ostensibly in the social interest and provides stability to the economic order.

The IISARMP works in tandem with a bevy of other policies related to governance of digital platforms, algorithms, and artificial intelligence. The most notable examples are the Personal Information Protection Law (PIPL), the Data Security Law, and the Cybersecurity Law. A final version of PIPL was released on 20 August of 2021 and became effective law several weeks later on 1 November. PIPL is "a special legislation on personal information protection… [that] contains the basic principles, requirements and related systems for the protection of personal information."[8]  It is therefore often described as a Chinese version of the EU’s GDPR given the deep parallels and similarities between them, however there are small differences. This legislation is expected to have a significant impact because it "solves the [current] problem of inadequate and scattered personal information protection legislation."

A central objective of PIPL is to clarify the rights of users over their personal information when interacting with internet-based services. The law introduces new protections against profiling and extends new rights for users to customize how their data is used by data processors. Examples include the ability to turn off targeting based on individual characteristics, request exclusion from automated decision-making, and the ability to provide user feedback. The bill also advances key changes to the control, ownership, and use of personal data. Through PIPL, individuals gain the rights to, "inquire about what personal data is being collected and stored by the data processor… to request a copy of their personal data, correct any inaccurate personal information, and delete their personal information when withdrawing consent or terminating the use of the product or service."[9]

Other relevant rights from PIPL come in the form of explicit duties placed on data processors. The requirement for processors to obtain user’s consent (to data collection) will encourage processors to introduce an 'opt-in' interface as opposed to an opt-out one. Relatedly, Article 15 stipulates that processors must "provide a convenient way to let the user withdraw their consent".[10] These strong consent-based rights are further enhanced by the fact that, under PIPL, processors cannot refuse service if a user denies or withdraws their consent, unless the service could not process without it. Finally, the law also provides safeguards against unwanted sharing of data with third-party entities by requiring processors to obtain additional and separate consent to do so. Additional consent is also required for the processing of 'sensitive personal information', which is a considerable step given that, "The scope of 'sensitive personal information' in the PIPL is much broader than in the GDPR," with, "financial information, transaction records, and location tracking are [being] regarded as sensitive personal information."

One of the most recent regulatory focuses of the Chinese government is 'deep fake' technologies and services. On 10 January 2023 China released is 'Deep Synthesis Provisions' to 'to strengthen its supervision over deep synthesis technologies and services'.[11] These provisions extend obligations on both generators and users of deep synthesis services, with the aim of promoting transparency, data and personal information protection, content management and labelling, and technical security. The 'comprehensive scope' of the deep synthesis provisions makes China a leader in the regulation of this kind of technology. As Kachra (2023) explains, "While the UK is also intending to ban the creation and dissemination of deepfake videos without consent, China’s law goes beyond this. The regulation creates rules for every stage of the process involved in the use of deepfakes, from creation to labelling to dissemination, leaving room for the potential suppression of organically captured content as well."

Provincial and local regulation

AI regulation and development has not only come from national legislatures and agencies. Provincial and local actors have started to put forward their own policies as well. A notable example is the 'Regulations on promoting artificial intelligence industry' passed in the Shenzhen Special Economic Zone. The purpose of the Shenzhen AI Regulation is to 'to promote the AI industry by encouraging governmental organizations to be the forerunners in utilizing related technology and increasing financial support for AI research in the city'.[12] These efforts align with the Shenzhen’s interest in being China’s leading site for AI and tech development. The city has pledged an investment of $108 billion over five years (2021-2025) to 'to reinforce its position as China’s innovation powerhouse'.[13] The regulations also contain guidelines for how public data is to be shared when supporting organisations and businesses. Other important elements of the policy include green-lighting trials of 'low risk' technologies and a call for establishing AI ethics council to 'develop safety standards and examine how the technology will affect things like employment, data protection, and other societal concerns'.[14]

Shanghai city has also been proactive in the realm of AI development and regulation. In September 2022 the provincial government passes a law entitled 'Shanghai Regulations on Promoting the Development of the AI Industry'. Government sources explain that the purpose of the bill is to 'promote the building and use of public computing resource platforms and provide public computing power support for AI technology and industrial development'.[15] Similar to the Shenzhen law, this policy contains many key components, including the introduction of a 'grading management and "sandbox" supervision' to help guide the testing process new technologies, establishing ethical norms and guidelines for AI research and development, and setting boundaries for toleration of ‘minor infractions to encourage exploration of scientific frontiers and inspiring innovation’.[16]

Development

The Chinese government has approached AI development with both great caution and enthusiasm. Whilst the risks are well-recognized by State leaders, China is taking a multi-faceted approach to support the development and growth of its AI industry. Key efforts include implementing a strong and clear regulatory framework, promoting AI infrastructure development and security, cultivating internal talent through increased R&D funding, and international collaboration.

State-led policy initiatives are playing a sizable role in the growth of China’s global standing in the AI race. One of the earliest testaments to China’s ambitions in this space was the 'Made in China 2025' initiative launched in 2015. It commenced a significant shift in China’s determination to be a major player in the technology space on a global level, rivalling Europe and especially the United States.[17] Key objectives of the plan included upgrading China’s internal manufacturing capabilities, a reduction in dependence on foreign technology, and fostering innovation across several industrial sectors (especially high-tech ones). The government no longer refers to the 'Made in China 2025' policy, but the effects of it are still observable and discussed.

The New Generation AI Development Plan released in 2017, discussed in the previous section, was a landmark policy that has served as a roadmap for China’s AI development strategy. The publication of this Plan demonstrates China’s unique long-termist approach to AI development that places great emphasis on the State as an active participant. The Plan stresses the importance of cross-sector collaboration and integration to reach key goals by 2030. State-led imitative is also observable in the 'Five-year plans' released by the Communist Party of China (CPC). These 'short and general guidelines' function as a 'central means of organizing policy in China, especially in the areas of environmental protection, education, and industrial policy'.[18] The most recent 14th plan (2021-2025) emphasised AI as central industrial sector for rapid technological innovation and economic growth.

Another important AI-related project unfolding in China are AI pilot zones. The Ministry of Science and Technology reports that China intends to build 'around 20 AI pilot zones by 2023'.[19] These zones will serve as investment hotbeds to 'to innovate a number of practical and effective policy tools, explore models integrating AI and economic development, gain experience that can be repeated and promoted, and build highlands for leading AI'.[20] These zones are marketed as opportunities for foreign investors to support the AI sector at the regional level. According to Wan Gang, president of the China Association for Science and Technology, 'will attract global innovation resources to deeply integrate with China's real economy, and constantly create a new trend of industrial development while becoming a new engine of economic and social development'.[21] As of May 2023, China has already established 18 zones. Each one is intended to address local technological needs and target different aspects of AI growth and development. 

The innovative development techniques by China also include the creation of regulatory sandboxes. The government has launched initiatives to develop these sandboxes so that AI technologies can be tested in real-world conditions whilst mitigating any potential risks. The Shanghai Regulations on Promoting the Development of the AI Industry includes were passed in part to 'further pave the way for sound and sustainable development of the AI technology through a grading management and "sandbox" supervision'.[22] The purpose of this regulation in the eyes of government officials is 'to provide sufficient space for companies to explore and test out their technologies.' Similar projects have been launched in Beijing, where the China Securities Regulatory Commission has 'launched the fintech innovation pilot projects in Beijing to promote the stable development of fintech in the securities and futures industry'.[23] The Director of the Institute of Finance and Banking in the Chinese Academy of Social Sciences, Yin Zhentao, has noted that the 'the fintech innovation pilot program is an important form of "regulatory sandbox", and it supports and encourages financial innovation, especially fintech innovations on the basis of keeping regulatory compliance and risks under control'.[24]

A variety of indicators suggest that China’s industrial strategies are paying off. The rise of the industry, with heavy state-backing, has been meteoric. The country has witnessed an explosion of economic activity in the AI sector, which has now reached a valuation of 578.4 billion yuan ($79.67 billion) in 2023, with the number of AI enterprises exceeding 4,000.[25] A white paper released by Deloitte emphatically notes that 'China has become one of the world's fastest growing countries in AI development.'[26] As of the second quarter of 2024, the "(AI) sector has nudged aside finance to top the country’s salary rankings."[27] These trends are expected to continue into the foreseeable future. A report released by the CCID Consulting expects the 'the market scale of China's AI industry will reach 1.73 trillion yuan by 2035, accounting for 30.6 percent of the world's total.'[28]

It also worth noting the geographical distribution of the AI industry across the country. Shenzen is currently regarded as the dominant hub of the AI development in China. This is not surprising given the efforts made by the city to attract AI entrepreneurs and businesses. In 2022, Shenzen enacted the 'first-ever local regulation dedicated to boosting the development of artificial intelligence (AI),' in an effort to 'super-charge its high-tech sector.'[29] The following year (2023), Shenzen local government released plans to 'build a digital platform to coordinate the deployment and use of computing resources for artificial intelligence projects in the city'.[30] The plan also includes the building of of an 'AI supercomputing network'. Beijing and Shanghai are also home to a vibrant AI sector. Beijing, being the home to many of China’s top universities, can draw on a strong talent pool of researchers and engineers. But other cities are making ambitious bids for relevance in the sector too. It has recently been reported that the AI "industry in China’s southern Guangdong province grew to 180 billion yuan (US$25 billion) in 2023."[31]

Whilst recent years have seen great AI prosperity in China, significant obstacles loom ahead. The growing geopolitical rift between Beijing and Washington threatens to cause a slowdown in the industry, and in fact it already is having such an affect. In addition to state resources, foreign private capital has been a key ingredient fuelling the rapid growth of the sector in China – especially venture capital from the United States. This source of investment, however, has been impacted by U.S.-China decoupling over geopolitical tensions: 'American venture capital, which has historically been the major drive of growth in China’s internet sector, has plummeted [in 2023]'.[32] The drying up of U.S. capital is the result of several reasons, many of which stem from the U.S. increasingly regarding China as a geopolitical foe instead of a mere competitor in the global economy.

The Economist, a proxy for the geopolitical sentiments of Western (NATO-aligned) governments, has been raising concerns around China’s growth in the AI sector at least since 2017, and with notable frequency ever since.[33] In recent years, a narrative has arisen around the so-called 'China threat'.[34] The Federal Bureau of Investigation bombastically claims that "the Chinese government is seeking to become the world’s greatest superpower through predatory lending and business practices, systematic theft of intellectual property, and brazen cyber intrusions."[35] AI development in China is thus viewed by American elites as a central part of the forthcoming (if not already present) competition between great powers in a multipolar world. The American Security Project, a hawkish neoconservative Washington DC think tank, puts it bluntly: "China’s aggressive pursuit of artificial intelligence (AI) harkens a future where bytes and algorithms challenge traditional battlefield superiority. If the U.S. government does not set guardrails now, American firms pursuing lucrative Chinese markets may vanguard the Chinese Communist Party’s transition to fifth-generation warfare."[36]

It is in this geopolitical context that the U.S. government has made AI investment in China riskier for U.S. companies and nationals. There are already strong restrictions in place that can make the movement of U.S. investment funding into Chinese companies an arduous task. Before Chinese AI companies can obtain funding from U.S. institutions, 'they need to have the appropriate corporate structure, offshore data storing solutions, and even foreign passports for their founders so the Silicon Valley investors won’t worry about violating U.S. restrictions'. Investors are concerned that further regulatory scrutiny is on the horizon. In June of 2023 'The House Select Committee on the Chinese Communist Party sent letters to four separate U.S. venture capital firms, including Qualcomm’s venture arm, expressing "serious concern" about their investments in Chinese tech startups.'[37] Such expressions by governing and legislating officials have cooled investor confidence in Chinese AI start-ups as a safe outlet for capital. Other issues abound like the unlikelihood of Chinese tech firms being listed on U.S. stock exchanges – which makes investments much less liquid – and the shortage of chips that will result from the ongoing chip war.[38]

Governance

Resistance to the deployment of AI in China has come from many different sources, which is yielding a complex dialectic of governance around this technology. China, as has been widely reported in the ‘Western’ press, has been a global leader in integrating digitalisation in to social and economic life. In some instances, the advancement of this high-tech movement has faced intense backlash. The widespread use of facial recognition technology across various social spaces – malls, schools, streets, etc., -- has generated concern and criticism from some quarters. There is the famous case of Hangzhou Safari Park, which required visitors to submit to facial recognition scanning to gain entrance. Professor Guo Bing of Zhejiang University famously sued the park, reportedly stating "I [filed this case] because I feel that not only my [privacy] rights are being infringed upon but those of many others".[39] As the Guardian explained, Bing’s efforts were guided by an effort to 'highlight its misuse of data gathered by the software'.

The problem of privacy continues to find its way into the Chinese courts. In 2020, The Beijing Internet Court made two major rulings on the legality of data practices by Chinese apps. In one of the cases, a plaintiff surnamed 'Huang' 'sued tech giant Tencent after she discovered that her account on the company’s e-reader app, WeChat Read, was automatically linked to her WeChat contacts, and that her virtual bookshelf was viewable to them without her consent.'[40] This court’s ruling provided a partial victory for Huang, finding that Tencent had infringed upon her 'personal information' and ultimately awarded her a small pay-out. However, the court equally argued that no privacy violation occurred, stating that access to her virtual bookshelf "can also meet the user’s expectation for knowledge sharing and cultural exchange." Tencent indicated to the press acceptance of the outcome. A similar case that year came before the Beijing Internet Court, with a plaintiff surnamed Ling suing Douyin for improper use of her data, resulted in the same outcome: the court ruled that Douyin had infringed upon her personal information but not violated her privacy. Bytedance, the parent company of Douyin, said they would appeal the ruling.

Western 'democracies' and their proxies have certainly been vocal in terms of their concerns regarding the rise of a so-called 'surveillance state' in China. One does not need to look very hard to find condemnatory news articles about the far-reaching surveillance powers of the Chinese government. The rise of a 'social credit system' at one point generated notable fear about the technocratic powers of the Chinese State in the minds of western citizens. This system promised to be 'a set of databases and initiatives that monitor and assess the trustworthiness of individuals, companies and government entities. Each entry is given a social credit score, with reward for those who have a high rating and punishments for those with low scores.'[41] In short, a giant algorithm would enforce black- and redlines against individuals based on their standing as a citizen. Regrettably, media-generated frenzy over this development caused serious misunderstandings about its true nature. As the Mercator Institute for China Studies explains, while such a system does exist, 'it is very different from what is imagined by many critics outside China'. They further explain how it all went wrong:

Some commentators seem to imagine that a magic algorithm draws from AI cameras and internet surveillance all over the country to calculate a score that determines everyone’s place in society. In reality, the SoCS is not the techno-dystopian nightmare we fear: it is lowly digitalized, highly fragmented, and primarily focuses on businesses. Most importantly, such a score simply does not exist… [And] The personal scoring initiatives that live on today serve only as positive incentives.[42]

There have been numerous protest movements or bouts of civil unrest demonstrating widespread dissatisfaction with algorithmic management systems. Drivers for ride-hail apps and couriers for delivery apps have repeatedly staged demonstrations and protests over poor working conditions and the harms of algorithmic management. On the 16th-18th of November, for example, a nationwide strike was conducted by drivers Huolala 'in large part over a new multi-factor ordering system on the platform'.[43] This new ordering system, drivers have argued, is reducing the price of trips and driving down incomes – and it is all the harder to know why because the algorithm is opaque. The China Labour Bulletin notes that in the Huolala situation, "[platform] Workers are up against an unclear and inhumane algorithm, a powerful company monopolising and setting standards in the industry, and a regulatory process that is at most a slap on the wrist."[44]

Much of the resistance to AITs in China come directly from mass-movements that ultimately require a response by regulators. A well-documented example of this is how gig workers have forced a national reckoning on the use of algorithms by major tech companies in the platform economy. In September 2020, a report called "Delivery Rider, Trapped in the System" went viral on social media, triggering the most heated discussion ever on possible algorithmic manipulation in China.[45] Workers who are managed and controlled by algorithmic systems are exposed to high risk during the course of their work, necessitating active regulation on AITs used to operate platform work.

Following that public outburst, in July of 2021 a guideline called 'Guiding Opinions on Protecting the Labour Rights and Interests of Workers under New Forms of Employment' (hereafter as 'Document No.56') was published by the Ministry of Human Resources and Social Security (MOHRSS) together with seven other departments. This was the first time the labour department responded to and addressed the issue concerning the way algorithms apply to the work field, despite the fact that the issue has existed for some time. Document No. 56 emphasises enhancing algorithmic transparency. The regulation says that those algorithms directly related to workers' rights and interests shall be fully asked for the opinions and suggestions of trade unions or workers' representatives. Subsequently, another more detailed document called 'Guidance on the implementation of the responsibility of online catering platforms to effectively safeguard the rights and interests of take-away food delivery personnel' was published by State Administration for Market Regulation (SAMR) with other government agencies, specified in the field of online catering. In accordance with the requirement for transparency in Document 56, the Guidance stipulates that algorithm rules affecting the vital interests of food delivery workers must be made public in advance so that food delivery workers, unions, and other parties can provide their advice.

We can see that the media, the courts, and social media campaigns are, in certain respects, putting pressure on companies and the state to adopt new policies that better protect and respond to the needs of data subjects. The role of trade unions is a much more complicated affair. This is in part because the All-China Federation of Trade Unions (ACFTU) is "China’s sole legally authorized central trade union, which comprises layers of unions at the regional and industrial levels."[46] The international trade union movement has been very critical of this fact, arguing that China ultimately lacks any significant or real trade union movement as the ACFTU is loyal to the CPC. The limits of the ACFTU are deeply evident when it comes to platform workers. In their major review of the Platform Economy, last updated April of 2023, the China Labour Bulletin laments that, "China's official union has failed to organise workers and initiate collective consultation in the industry, and platform workers continue to lack adequate representation."[47] It is for this reason they recommend that 'the official trade union take a greater role in protecting platform workers' rights and interests by creating platform industry sub-divisions within its official structure'. This would allow for union members, or workers’ to directly participate in union activities, and 'establish a level of accountability.'[48]

There is some evidence, however, that the government is creating space for a more powerful and active labour movement. This shift can be linked to widespread pressure by labour activists, but also the policy paradigm shift commenced with Xi’s 'Common Prosperity' agenda.[49] The recent regulatory initiatives undertaken by the Chinese government described above, include efforts to improve workplace conditions, protections, and voice to contest AIT abuse and overreach. In July 2021, China’s Ministry of Human Resources and Social Security published a document titled Guiding Opinions on Protecting Labor and Social Security Rights and Interests of Workers Engaged in New Forms of Employment that ‘specifically target[s] the rights of platform workers’.[50] A key innovation in these Guidelines is the mandate that 'employers enter into contracts with workers even when their relationship does not meet all of the requirements for a legal "labour relationship"' and also the emphasis the Guidelines place on the role of labour unions.[51] This is especially evident in Article 10, which makes reference to not only trade unions but 'workers’ representative'. This underscores a new flexibility and openness to the claims of workers directly, likely an outflow from the vocal criticism by workers of platform companies and the ACTFU.

Another key development is the recent revision of China's Trade Union Law. This law is responsible for setting out the trade union and collective bargaining rights prescribed to workers, including the right to organise, oversight of termination and layoffs, health and safety requirements, autonomy in union activities, and so on. The Trade Union Law was amended in 24 December 2021, with the revisions take effect 1 January 2022. One of the key changes to the law was the addition of the following clause to Article 3: Trade unions adapt to the development and changes of enterprise organisation forms, workforce structure, labour relations, employment forms, etc., and safeguard workers' rights to join and organise trade unions according to law. This constitutes a potential major step forward for platform workers as ‘It is generally believed that this new clause clarifies gig workers' right to join and organise trade unions'.[52] However, as noted in the regulation, there is ambiguity about what this change means, and it may take time for the courts to establish its significance. A further example of how labour activism is changing the trade union landscape is Article 10 of Document No. 56. This guideline requires that digital labour platforms explicitly consult workers on how algorithms should be designed and used when those algorithms directly affect labour rights. This may lead to the co-production or co-determination of algorithmic rules that better track the interests of workers.