China’s ruling Communist Party is using artificial intelligence to increase surveillance and control over its 1.4 billion people, with the technology becoming even more pervasive in daily life, predicting public demonstrations and monitoring the mood of prison inmates, according to a new report.
Many of these systems are already well-documented, from the army of online censors maintaining the Great Firewall to the ubiquitous surveillance cameras on nearly every street and block in urban China.
But a report released by the Australian Strategic Policy Institute (ASPI) on Monday details how government AI tools used to “automate censorship, strengthen surveillance and pre-emptively suppress dissent” have become more sophisticated over the past two years, against a backdrop of increasing technology competition between the US and China.
“China is leveraging AI to make existing control systems much more efficient and intrusive. AI will allow the CCP (Chinese Communist Party) to monitor more people more closely with less effort,” said Nathan Attrill, co-author of the report and senior China analyst at ASPI, which receives some funding from the Australian government and other foreign governments.
“In reality, AI has become the backbone of broader, more predictive forms of authoritarian control.”
The authors added that the implications would be wide-ranging and deep, allowing the Chinese government greater control in policing its population and managing the flow of information, as well as increasing its power abroad as a global exporter of surveillance technology.
Despite US efforts to limit the supply of high-performance AI chips to China, the Chinese government has invested hundreds of billions of dollars in AI-related businesses and made significant advances in research and development.
The public is also embracing this technology. A 2024 survey by global research group IPSOS found that Chinese respondents were far more excited and optimistic about AI than other respondents across 32 countries.
Even Chinese leader Xi Jinping has emphasized the importance of AI in the country’s evolving internet policy. According to Chinese state media, in a November meeting with Chinese Communist Party leaders, he emphasized that AI “presents challenges to the governance of cyberspace while also providing new means of support,” which the ASPI report argues is a euphemism for maintaining the regime’s power and stability.
ASPI’s findings are not entirely new. Other researchers and institutes around the world have previously published similar reports and warnings. Chinese leaders have spoken openly about their AI ambitions, some of which are shared by other countries. And it’s still not the national standard. Local governments in large urban hubs with existing digital infrastructure, such as Beijing and Shanghai, are experimenting with AI in ways that rural areas and smaller cities are not yet able to do.
But “many of the government’s intentions and policies are now becoming reality,” said Xiao Zhang, a researcher on internet freedom at the University of California, Berkeley.
He added, “This report provides clear indicators that China is moving towards[leveraging AI nationally]. …Those things will be implemented as soon as the digital infrastructure is ready.”
AI is currently being used in some places, including police enforcement, court proceedings and prison operations, and the report argues that the technology could eventually be integrated into every stage of China’s already opaque criminal justice system.
Surveillance begins with China’s vast network of surveillance cameras. There are no comprehensive statistics on the number of cameras in the country, but the report estimates there are up to 600 million cameras across China. This equates to approximately 3 cameras for every 7 people.
As in many other countries, these cameras are increasingly equipped with AI features such as facial recognition and location tracking. For example, the report found that documents from one district in Shanghai detail plans for AI-powered cameras and drones to “automatically detect and intelligently enforce laws,” including the possibility of alerting police to crowd gatherings.
According to the report, China’s Supreme Court has asked all courts to “develop competent artificial intelligence systems by 2025” that can be used in various legal procedures, including trials and administrative affairs. As an example, Shanghai’s AI system can reportedly recommend whether judges and prosecutors should arrest criminal suspects or defendants or give them suspended sentences.
Finally, there is also a push toward “smart prisons,” where AI tools can track prisoners’ locations and movements. In one prison, facial recognition cameras monitored inmates’ facial expressions and flagged an intervention if a prisoner appeared angry. At a drug rehabilitation center, prisoners received AI-assisted therapy via virtual reality (VR) headsets.
“Defendants arrested through AI-based surveillance and tried in AI-assisted courts could be sentenced based on the AI system’s recommendations for ‘smart prisons’ incorporating a wide range of smart technologies,” the report states.
China’s State Council Press Office and the Ministry of Justice did not respond to CNN’s requests for comment. They have previously criticized ASPI for receiving funding from U.S. government agencies and argued that ASPI has “no credibility.”
These smart technologies could help prevent crime and make China’s cities much safer, Xiao acknowledged, but “because of the political system, the same technologies can and have been used (for) political persecution.”
China’s court system, which is at the mercy of the Chinese Communist Party, already boasts a conviction rate of over 99%.
Mr. Xiao pointed to several vulnerable groups that could be further targeted, including religious and ethnic minorities like the Uyghurs and political dissidents who have long faced government repression.
According to the report, Chinese companies are also currently working on developing large-scale language models (LLMs) for minority languages such as Uighur, Tibetan, Mongolian, and Korean, with support and funding from the central government, to improve “monitoring and control of communications in these languages.”
According to Xiao and his report, these LLMs could be used to monitor what minority communities post and share and manipulate the information they receive.
The report also highlights the role of China’s largest technology companies, calling them “key enablers and enforcers of the Chinese Communist Party’s online content censorship policies.”
These companies, which have always been required to abide by central government content regulations, are now key players in developing and selling censorship technology to small and medium-sized businesses across the country, and sometimes cooperating with authorities in criminal cases, the report said.
For example, TikTok’s parent company ByteDance censors content on Douyin, a version of the app used primarily in China, blocking or downvoting politically sensitive content.
Social media and gaming giant Tencent uses AI to monitor user behavior and assign “risk scores” based on online activity, including penalties for violations across social media, chat groups and other communication platforms, the report said.
Search engine Baidu sells a number of content moderation tools and has worked with government agencies in more than 100 criminal cases, mainly related to fraud and cybercrime, the report said.
“Online, AI enables real-time censorship and opinion shaping, with platforms using automated moderation, sentiment analysis, and recommendation algorithms to down-rank criticism and promote narratives tailored to political parties,” said Attrile, a co-author of the report.
CNN has reached out to all three companies for comment.
The report warns that the growth of China’s ecosystem of AI surveillance and censorship tools, which small and medium-sized enterprises alike are developing in-house, is having global implications, with other authoritarian states such as Iran and Saudi Arabia also using AI to monitor their citizens.
“Recent LLMs in China are dominated by the open weight model. This means that many other countries, including their own companies and research institutes, may use the Chinese model because it is cheap and free,” Xiao said.
But “with these models, you’re basically sitting on that platform,” he added. “It involves censorship, surveillance, control and influence.”
