[ARTICLE] Making sense of China’s paradoxical AI strategy

28 October 2021

By Guillaume Guinard

While the latest Facebook scandal and a growing concern for digital sovereignty have made platform regulation a key priority for both US and EU regulators, Chinese authorities have been cracking down on national tech giants during the past year – particularly in relation to their use and collection of data. This has recently culminated in a set of ethical guidelines for AI, which establish requirements for transparency and accountability at odds with the business models of major Chinese tech firms. This could seem paradoxical given the Chinese ambition to become the AI world leader by 2030. China’s goals in terms of AI development are indeed distinct, as they prioritise healthcare and military technology over entertainment. At the same time, the government’s sovereignty over its citizens is challenged by its tech giants in a similar way to those of the US and EU. As such, despite the numerous lawsuits filed against them, China’s approach to regulating AI-powered platforms shouldn’t be dichotomised with that of the West. In both cases, states seek to establish a relationship of cooperation and regulatory compromise with digital behemoths.

In 2017, China announced its plans to become the world leader in AI by 2030. In order to clarify its intentions, the government established a three-step roadmap. First, it would develop standards, policies and codes of ethics for the development of AI by 2020. Second, it would achieve a major breakthrough leading to industrial upgrading and economic transformation by 2025. Third, the industry would reach 1 trillion yuan (150 billion USD) by 2025-30. At the time of the announcement, Chinese tech giants seemed like the driving force behind these ambitions, with Baidu’s research centres in Silicon Valley exploring the possibilities of self-driving cars, and Alibaba heavily investing in AI development in areas ranging from shopping to healthcare.

How have they been doing so far? Well enough for the government’s ambitions to be taken seriously, to say the least. In 2019, a report from the Center for Data Innovation placed China third in the ‘AI Race’, behind the EU and the US, but noted that it was growing faster than both its competitors. An ex-software chief from the Pentagon even told Reuters recently that China has already won the AI battle with the US, citing among other things the refusal of giants like Google to cooperate with the US government.

It could seem paradoxical, then, that the first step of China’s roadmap has involved a year-long crackdown on its own tech giants, specifically over their use and reliance on AI. Especially considering the high cost of these operations: regulatory action against ride-hailing app Didi caused the value of its stocks to plunge shortly after they had reached an overall valuation of 100 billion USD in the US market.

It is tempting, at first, to characterise these acts as an attempt of the Chinese Communist Party to establish its dominance over the private sector as major Chinese firms attempt to go global. Then-White House Communications Director Anthony Scaramucci went so far as to label the Didi IPO incident a ‘direct assault on global capitalism.’ It is also possible to see China’s recently-revealed AI ethics guidelines as an obstacle to the innovative breakthrough the government strives for. Its requirement for humans to have the right to opt out of an interaction with an AI system is at odds with the business models of many companies which use and invest in AI in the advertising industry. Furthermore, tighter requirements regarding ‘self-censorship’ and ‘integration of ethics into every stage of research and development’ could impose higher financial costs which weed out small businesses from the market, as well as slow down these operations, making it harder for Chinese businesses to keep up with the pace of innovation of their international competitors. Lastly, the guidelines’ requirements of transparency and the ability for authorities to intervene in AI systems could signify a total ban on elaborate algorithms whose internal operations are too complex for humans to comprehend and thus intervene in.

There is some truth to these views. In particular, it is clear that the CCP is seeking to use and regulate AI in order to further its political goals and assert its control over Chinese society. In June 2021, a Tencent and Alibaba-backed tech startup was put under investigation by the Cyberspace Administration of China for a social media post implicitly referring to the anniversary of the 1989 Tiananmen Square massacre. Additionally, two years ago, leaked documents showed how the Chinese government is investing in and developing policies surrounding facial recognition technologies in order to optimise surveillance over the Muslim ethnic minority in its Xinjiang province – a flagrant breach of its own recently released ethical standards regarding algorithmic transparency, and how AI should not be used to harm minorities.

Nevertheless, in order to fully understand China’s ambition to become the global AI leader, it is necessary to paint a more nuanced portrait of its incentives and action plan. These can be summarised in three core aspects: a targeted approach, sovereignty over citizen data, and fostering self-regulation.

China’s vision of AI development is targeted and revolves around local ecosystems

Dr. Kai Fu Lee, former head of Google China, referred to the Chinese model of innovation as one that focuses on ‘fusion and speed’ over breakthrough technologies, which seeks to optimise the use of current technology in localised areas. Concretely, this means taking advantage of China’s local ecosystems and prioritising experimentation. Contrary to Western narratives surrounding the crackdown on big tech, China’s approach is not top-down. National guidelines are established by a consortium of universities, local government, and private entities, and their implementation relies on experimentation by local ‘AI Towns’, such as Hangzhou. Started in 2017, the ‘Hangzhou AI Town’ seeks to link together e-commerce company Alibaba, Zhejiang University, graduates returning from overseas and local businesses in an AI cluster, connected to larger infrastructures of science in technology in the city. The goal of the AI town is to cultivate knowledge spillovers and cooperation, in line with China’s broader bet on open-source software in order to boost innovation. These multiple local ecosystems are encouraged to experiment before implementing guidelines, in a fashion that European legal frameworks would not allow (in order to protect citizens). Finally, the concerns of citizens themselves have led to regulatory changes, among which are concerns over facial recognition systems that led to several cities restricting the use of the technology.

Moreover, China is prioritising leadership in AI in fields that it deems to be in line with its societal priorities. For instance, in recent years the country has seen rapid development in the field of AI for healthcare. Examples include AI doctor chatbots to help people in remote areas via telemedicine, and deep learning applied to medical image processing. Some AI tools developed by Chinese companies in the context of the global pandemic are now used by the NHS in the UK. It is also worth noting that Chinese state-owned companies, as well as companies with ties to the government, have been investing billions of dollars in US military tech startups, while at the same time dialling back investments in entertainment. The crackdown on tech giants in the media and video game industry is hence unsurprising, as it follows the same trend. China’s strategy seeks to develop AI for purposes beyond economic benefit, and therefore does not include investment in or protection for entertainment.

Sovereignty is a key priority within the country’s strategy

Indeed, much like the EU, China is concerned about the government’s sovereignty over the data of its own citizens – especially given that American laws require companies to share this data with the US government upon request. This constitutes another reason for the promotion of open source software under China’s AI strategy, as it seeks to circumvent and decrease reliance on American proprietary software.

Moreover, the aforementioned crackdown on the ride-hailing app Didi occurred shortly after the company entered the US stock market, and it is rumoured that the company had been in touch with the US government over data usage. While the concerns over misuse of customer data were likely well-founded, the timing of the investigation strongly suggests it was the collusion with the US government that caused Chinese authorities to act so swiftly and authoritatively. An editorial published in the state-backed Global Times explains that Didi, with 80% of the Chinese ride-hailing market, holds sensitive information about personal travel and habits. It said the government won’t let internet giants ‘become rule-makers of data collection and usage,’ adding that ‘the standards must be in the hands of the government.’

This suggests that the Chinese government is not merely seeking to establish its sovereignty over foreign nations, but also over its own private sector. Years of lax regulations to favour rapid growth have enabled tech industry actors to grow to a size where they have become not merely too big to fail, but also big enough to threaten the state’s sovereignty. While Beijing has more power to investigate and restrict tech giants over vague pretences than European and American authorities, and can do so with greater speed, it cannot simply quash them without risking severe societal disruption.

The power of its tech giants means Beijing must foster self-regulation.

Throughout the past year, different Chinese government agencies have issued fines for misuse of data, most notably local governments. For instance, the Shanghai Municipal Administration for Market Regulation recently imposed a 500,000 yuan fine – the equivalent of about $78,000 – against Alibaba-owned delivery app Ele.me for breaking China’s price and food safety laws. Another example is the Chinese cyber regulator, which gave short video and job recruitment apps three weeks to fix their illegal use and collection of personal data. Angela Zhang, an associate professor of law at the University of Hong Kong, told the Wall Street Journal that the government’s aim is to ‘nudge firms to comply with the regulatory demand without formal intervention’.

In contrast to the cyber regulator’s desire to avoid formal intervention, major fines have been issued by the Chinese antitrust authority towards almost every big platform company, including Tencent (video games and music), Didi, Baidu (search engine) and ByteDance (social media, most notably TikTok). Alibaba was hit first and hardest, receiving fines to the value of a record 2.8 billion USD in April 2021. These fines were issued for a range of reasons, including failing to disclose mergers and engaging in misleading marketing tactics. Importantly, many of the sanctioned deals occurred prior to 2018, before the watchdog even existed.

Through this flashy and sweeping action, Beijing is sending two important messages to these firms. First, no platform is large enough to escape government control. Second, getting away with unethical behaviour because regulation is lagging behind industry practices doesn’t rule out being punished for it at some point later on. Beijing seeks to get rid of the sense of impunity which these platforms had enjoyed so far, in order to establish the prerequisite willingness to cooperate necessary for self-regulation.

Indeed, while it is tempting to dichotomise China’s strong ‘authoritarian’ approach to platform regulation and the Western context, where platforms are asking to be regulated, the reliance of government institutions on these platforms and the complexity of their products means that neither China nor the West can claim to be fully in control. This is why the recently released ethical guidelines for AI emphasise transparency and oversight, but also ‘self-discipline’. Not only are tech giants expected to govern themselves; they also play a big role in the development of the national authorities’ regulations. Executives from e-commerce giant JD.com and facial recognition startup Megvii are part of the AI governance committee of the MOST (Ministry of Science and Technology), which authored these guidelines. Moreover, Baidu and Tencent have submitted proposals on AI ethics directly to China’s leadership, and private companies are actively involved in the implementation of the guidelines in the AI towns mentioned earlier.

It is too soon to know whether the Chinese authorities’ sweeping actions have convinced companies to implement rigorous self-regulation of AI according to strict ethical and moral values. Antitrust lawsuits have so far proven effective in encouraging tech giants to signal their virtue by gifting large sums of money towards social aid, with Tencent pledging 7.7 billion USD to help governmental wealth redistribution efforts. Despite the power claimed by big tech in the negotiations surrounding AI regulation, the CCP seems to have managed to keep the upper hand.

Guillaume Guinard is a research assistant at the Digital, Governance and Sovereignty Chair, a master’s student in Public Policy at Sciences Po Paris and a Philosophy graduate of Glasgow University.