Matthew J. Dagher-Margosian
Executive Summary:
- Xi Jinping is unequivocal that US-China AI cooperation is contingent on Western AI technology flowing into the People’s Republic of China (PRC).
- Broken promises from past US-China technology transfers suggest the PRC will once again use US technology to strengthen censorship and surveillance domestically and then export these technologies to aid other authoritarian regimes.
- The emerging framework of “AI sovereignty,” born out of a previous framework for “cyber sovereignty,” emphasizes a nation’s right to regulate AI according to its own values and interests. The PRC positions itself as a leader in this discourse.
- The PRC is lobbying for artificial intelligence (AI) sovereignty globally and is already exporting AI technologies to other authoritarian regimes, raising concerns about the potential impact on global democracy and human rights.
For the past decade, the priorities of “economic security, national security, and security in other areas” have driven Xi Jinping’s pursuit of advanced technologies (Xinhua, June 9, 2014). The government of the People’s Republic of China (PRC) has launched countless policies to achieve “self-reliance” in technologies like software, semiconductors, mobile phone operating systems, and robotics (Quishi, July 31, 2023; STCN, November 14, 2023). For Xi, self-reliance is predicated on mistrust of the West. He has previously warned that “Western countries believe that the master will starve if he passes on his knowledge to his apprentice” and argued that the PRC “must focus on our own innovation” (People.cn, April 19, 2018). This has led to Western IT equipment, iPhones, and Microsoft operating systems being replaced with domestic technology across the PRC (WSJ, March 7).
The Chinese Communist Party (CCP) seeks a leading role in shaping global norms for artificial intelligence (AI). It is also calling for a United Nations AI governance body and has launched the Global AI Governance Initiative to demand global “information exchange and technological cooperation on the governance of AI” (MFA, March 7; October 20, 2023). The Party has rolled out the red carpet for those at the forefront of Western-led AI efforts. Former Minister of Foreign Affairs Qin Gang (秦刚), Ministry of Finance and Commerce Head Wang Wentao (王文涛), and Xi himself have personally discussed US-China cooperation on AI with leading tech figures, including Elon Musk, Bill Gates, Tim Cook, Synopsis CEO Sassine Ghazi, and Microsoft President Brad Smith (MOFCOM: March 25; December 6, 2023; March 3, 2023; MFA, May 30, 2023).
US business leaders have responded to the PRC’s outreach. Speaking at a Beijing AI conference, OpenAI CEO Sam Altman cited nuclear weapons as a precedent for global cooperation on mitigating existential risks (Global Times, July 10, 2023). Elon Musk has called for US-China AI cooperation as being in the interest of “team humanity” (Bloomberg, July 12, 2023). Microsoft and Google heads Satya Nadella and Sundar Pichai have also voiced support for bilateral cooperation (BBC Chinese, October 7, 2019; The Paper, November 17, 2023). Meanwhile, Microsoft, OpenAI, and Apple have begun collaborating with PRC AI researchers as well as with domestic companies like Baidu and Edianyun on the PRC’s domestic AI efforts (Financial Times, January 11; Global Times, March 26; 36kr, February 25).
US President Joe Biden acknowledged a need for more US-China dialogue on AI in his remarks to Xi at the APEC forum last November, as well as during a call in early April (People’s Daily, April 2; MFA, November 16, 2023). Their respective nations were among 28 to sign the 2023 Bletchley Declaration, pledging international cooperation on AI safety (Gov.uk, November 1, 2023). And this January, National Security Advisor Jake Sullivan met with Minister of Foreign Affairs Wang Yi (王毅) to discuss holding the first US-China bilateral meeting on AI risk (MFA, January 27). Director of the White House Office of Science and Technology Policy Arati Prabhakar has also recently alluded to future US-China AI dialogues (Financial Times, January 24).
PRC Demands for AI Cooperation
Major impasses remain for US-China AI collaboration. The Biden administration has implemented significant restrictions on the technology underpinning AI development through strict export controls on semiconductors to the PRC. It has also demanded that other countries follow suit (BIS, November 6, 2023). These controls are in addition to restrictions on exports of chipmaking equipment, potential future sanctions on Huawei semiconductors, and restricting PRC access to AI compute via cloud service providers (WSJ, July 4th, 2023; SCMP, Jan 27).
The CCP has been clear that there will be no substantial US-China cooperation over AI governance while the Biden administration blocks access to AI technology. At his 2024 World Economic Forum keynote, Premier Li Qiang (李强) stated that controls over advanced technologies like AI cannot “become a means to restrict or contain other countries” (Caixin, January 16). Wang Yi has also directly critiqued US policy, stating that “small courtyards and high walls” around AI are “historic mistakes” (MFA, March 7). [1] Meanwhile, Xi has begun to directly pressure US allies such as the Netherlands to provide the PRC with semiconductors lithography equipment, asserting that “no force can stop China’s scientific and technological development and progress” (AP News, March 27; Xinhua, March 27).
US IT and PRC Democracy: A Failed Gambit
History suggests that the PRC’s desire for collaboration on AI and its demands for AI technologies are not as genuine as they may seem. In the late 1990s, President Bill Clinton was convinced that providing the CCP with the transformative technology of that era—the Internet—would lead the Chinese populace to reject authoritarianism, famously proclaiming that controlling the Internet would be like trying to “nail Jell-O to a wall” (New York Times, April 4, 2000). In his dialogues with Clinton, CCP leader Jiang Zemin also seemed to genuinely embrace Western technology and expertise. In 1997, Jiang toured the United States to lobby for the PRC’s inclusion in the WTO and met with Cisco and Apple. (CISCO, September 21, 1998; Washington Post, October 29, 1997).
US media portrayed Jiang as cheerleading Western technology, but CCP documents tell another story. As far back as 1983, Jiang framed Western technology as a way to “accelerate the development” of the PRC’s technological “self-reliance” (Xinhua, April 22, 2009). In a 1999 speech, Jiang emphasized again that the PRC could not “blindly rely on foreign ready-made technologies” (MOST, December 22, 2006). Jiang’s approach was to offer US tech companies market access, allowing the PRC to “digest, absorb, improve, and innovate imported technologies to facilitate development of China’s technological advantages and enhance China’s ability to be self-reliant” (Xinhua, April 27, 2006).
Clinton’s gambit that the internet would conquer the CCP instead saw the CCP transform the internet into the greatest technological tool for surveillance, censorship, and authoritarianism in the PRC’s history (The Intercept, February 18, 2021; The China Project, August 19, 2020; ZDNET, October 24, 2016; Wired, May 20, 2000). Western tech firms whose knowledge helped build the CCP’s surveillance and censorship capacities have increasingly found themselves replaced by domestic alternatives (China Daily, September 24, 2015; DCD, August 15, 2019).
Betrayal and the Birth of Cyber Sovereignty
The PRC has now developed an ideological framework of “cyber sovereignty” to openly rebuke US “hegemony” over AI and other advanced technologies (Global Times, January 28; Global Times, June 29, 2023). This alternative framework for governing technology was first developed between 1994 and 2005 when PRC academics began debating cyberspace, sovereignty, and overcoming “digital colonialism” between China and Western powers (Oxford University Press, Oct 2022). These debates turned into policy when, in 2010, a State Council white paper declared the PRC’s domestic internet technology and networks under the jurisdiction of “Chinese sovereignty” (SCIO, June 8, 2010).
Xi Jinping later codified this policy framework as “cyber sovereignty (网络主权)”—the notion that both democratic and authoritarian countries have the right to “independently choose their own path of cyber development, model of cyber regulation, and Internet public policies” (FMPRC, December 16, 2015). [2] Cyber sovereignty as a framework for governing technology, has provided the CCP with a ready-made “kit” to export technology, expertise, regulations, and ideology abroad for authoritarian use (Qiushi, September 16, 2018; Pacific Forum, July 2019). This “kit” further assists authoritarian nations in their justification for rejecting the Western “hegemony” of democratic and open technological governance (World Internet Forum, November 9, 2023).
The PRC has worked since to form technological alliances with other authoritarian nations (Xinhua, May 9, 2015). In 2015, Russia and the PRC signed a groundbreaking pact to develop “universal” cyberspace norms (Xinhua, May 9, 2015). In 2017, the PRC signed the “One Belt, One Road” (Belt and Road Initiative; BRI) Digital Economy International Cooperation Initiative with Laos, Saudi Arabia, Serbia, Thailand, Turkey, and the UAE, with member countries pledging to “respect cyber sovereignty” (CAC, May 11, 2015). Cyber sovereignty was also a key requirement when the PRC and fourteen African nations signed the China-Africa Initiative to Work Together to Build a Community with a Shared Future in Cyberspace (CAC, August 25, 2021).
The CCP has also sought to install cyber sovereignty as a core value within global organizations. In 2015, a PRC-led member group within the UN proposed including cyber sovereignty as part of the International Code of Conduct for Information Security (CCDCOE, February 2015). With CCP support, Russia formed the “Open-Ended Working Group” (OEWG) in 2018 to oppose the American-led Group of Government Experts (GGE) and promote cyber sovereignty over democratic applications of tech governance. The lack of consensus between the GGE and the OEWG has led to an unresolved split regarding the UN’s approach to global internet governance (Just Security, July 16, 2021).
PRC efforts to spread cyber sovereignty are also backed by state funding and person-to-person transfers of knowledge. Officials, including the Great Firewall’s “architect” Fang Bingxing (方滨兴), directly consulted with Russian counterparts on censorship and surveillance (Eurozine, February 21, 2017). Similarly, Zimbabwe, Uganda, and Zambia have studied or adopted the PRC’s cyber sovereignty principles or been given direct technical assistance (Veritas Zimbabwe, March 11, 2022; WSJ, August 15, 2019; Foreign Policy, July 24, 2018; Sina, February 10, 2015). In addition, the PRC has loaned billions of dollars to companies like Huawei and Hikvision through government-run policy banks (IPVM, August 18, 2016). These state-approved loans have been traced to exports of surveillance technology and IT to authoritarian systems or weak democracies like Pakistan, the Ivory Coast, Mongolia, the Philippines, Belarus, and Serbia (Boston University Global Development Policy Center, accessed April 3; Aiddata, 2021; AP News, October 17, 2019).
Defining AI Sovereignty: Infrastructure and ‘Chinese Characteristics’
An emerging framework of “AI sovereignty,” or “sovereign AI” is evolving from the existing framework of cyber-sovereignty. This can be traced through public policy debates (Carnegie Institute, July 10, 2023). As far back as 2018, academics were predicting that China would need to participate in global governance due to the “sovereign costs” of AI. (Sohu, September 18, 2018). In 2020, the term “AI sovereignty (人工智能主权)” was defined by scholars Zhao Jun (赵骏) and Li Wanzhen (李婉贞) as the right for countries to internally regulate AI as they see fit (Journal of Zhejiang University, 2020). A more expansive framework later defined it as a country’s ability to regulate AI training data, how users engage with AI, and the transparency of AI algorithms (Journal of Shanghai University of Finance and Economics, 2022). These concepts draw explicitly on existing Chinese legal frameworks around cyber sovereignty, data sovereignty, and algorithmic transparency. [3] [4]
Related to “AI sovereignty,” Nvidia’s CEO Jensen Huang (黄仁勋) has advanced the idea of “sovereign AI (主权人工智能)” as a nation state’s ability to “develop AI infrastructure” while “protecting its culture and language” (Nvidia.cn, February 19; Sina, March 11). [5] Academics within the PRC have taken Huang’s definition to theorize that to achieve sovereign AI, countries should establish “local AI systems and infrastructure to meet their own specific production needs, control their own data and intelligence security, and prevent other countries’ ‘AI intrusions.’” Failure to establish sovereign AI would leave the PRC vulnerable to “digital colonization (数字殖民)” and “the penetration of Western ideological trends” (Outlook Think Tank, April 9). [6]
Within this discourse, “Chinese values (中国价值观)” are seen as crucial for the CCP’s control of sovereign AI. Top AI researchers like Professor Yao Xin (姚新) of Lingnan University, has advocated for domestic AI to reflect “Chinese Values” (Tencent Research Institute, March 21). Yan Kunru (闫坤如) of Shanghai University and a chief AI ethics expert at the National Social Science Foundation, has gone further, advocating that the PRC “must develop AI with Chinese characteristics,” adding that “we cannot be influenced by Western values” (Tencent Research Institute, March 21). Op-eds in state media have also emphasized that “there is no such thing as technology neutrality … [AI] has values, generated content has a direction, and AI products have a stance” (People’s Daily, October 29, 2023).
State media have further expanded on AI infrastructure and sovereignty. Li Zhiqi (李志起), director of the Beijing Chinese People’s Political Consultative Conference’s economic committee, has stated that “sovereign AI infrastructure (主权AI基础设施)” is a national security necessity and is “based in the localization of computing power” (Global Times, February 24). Tian Feng (田丰), founding dean of the SenseTime Intelligent Industry Research Institute, has emphasized that countries must protect their AI infrastructure and that “Artificial Intelligence is national sovereignty” (163.com, March 28).
Turning ‘AI Sovereignty’ into AI Policy
The thinking behind “AI sovereignty” or “sovereign AI,” although not explicitly mentioned in policy documents, is the emerging framework for the CCP’s governance. [7] Former Vice Minister of Foreign Affairs Fu Ying (傅莹) cautioned as far back as 2019 that the PRC must consider how AI “will inevitably impact the relationship of sovereign equality between large and small, strong and weak countries” (Sina, April 12, 2019). Minister of the Ministry of Science and Technology Wang Zhigang (王志刚) has warned of several “gaps” between the PRC and leading AI powers in areas such as chips, talent, and infrastructure (QStheory, August 5, 2019). To address weaknesses in the PRC’s sovereign AI, Premier Li Qiang has spoken with experts about the PRC’s lack of AI “self-reliance” and dependence on foreign open-source language models (CCTV, March 13). These same researchers expressed to Li their concerns over the immaturity of PRC semiconductors (SCMP, March 15).
Party research supports Wang’s warnings, with the China Academy of Labor and Social Security Sciences calculating the PRC needs at least 300,000 more AI researchers to meet current demands (Caixin, April 14, 2023). Other concerns that have been highlighted by domestic researchers include the country’s shortage of high-quality data necessary for training Large Language Models (LLMs) compared to Western nations (Tencent Research Institute, November 16, 2023; People’s Daily, October 29, 2023). Finally, flaws in China’s computing power capacity, have been cited as “fragmented” and “not conducive” to developing AI (SCMP, March 23).
The PRC’s existing AI laws also reflect the emerging AI sovereignty/sovereign AI discourse. The laws on generative AI from the Cyberspace Administration of China (CAC) explicitly call for AI developed in the PRC to “adhere to the correct political direction, public opinion, and values” and not “disrupt economic and social order” (CAC, November 25, 2022; CAC, April 11, 2023). Chinese Academy of Sciences (CAS) academics have drafted a hypothetical “National AI Law” calling on AI technologies to “adhere to the core socialist values, not incite subversion of state power, overthrow the socialist system, endanger national security and interests, damage the national image, incite secession, undermine national unity and social stability” (Data Law Alliance, August 16, 2023).
Exporting AI Sovereignty
Globally, the PRC is lobbying for AI sovereignty in both other countries and global organizations, just as it did for cyber sovereignty. CAC and MFA officials have publicly affirmed AI sovereignty as a key position for cooperation with the PRC (Huanqiu, November 25, 2023; MFA, December 21, 2020). The PRC’s UN ambassador Zhang Jun (张军) has also called for countries to establish “AI governance systems that are in line with their own national conditions” (PRC Mission to the UN, July 18, 2023).
In 2022, PRC representatives submitted the “Strengthening the Ethical Governance of AI” position paper to the UN. This paper requested flexible AI ethics based on “national conditions” and respect for “different countries’ AI governance” (MFA, November 17, 2022). Most explicitly, the PRC’s Global AI Governance Initiative states that global AI technology should respect “national sovereignty and strictly abide by [countries’] laws when providing [countries] with AI products and services” (MFA, October 20, 2023).
PRC companies lead the global export of AI facial recognition technologies, with nearly half of these exports heading to autocracies or weak democracies (Brookings, December 24, 2022). The China AI Exports Database (CAIED) recently found that the majority of the PRC’s AI exports are concentrated in non-democratic BRI partner countries, while other research shows that AI technology from the PRC undermined anti-CCP candidates in Taiwan’s elections (RAND, December 11, 2023; RAND, September 7, 2023; Microsoft, April 4).
Conclusion
Xi Jinping is unequivocal that US-China AI cooperation is contingent on Western AI technology flowing into the PRC (SCMP, May 26, 2023). But the broken promises of past US-China technology transfers should give the Biden Administration—and figures like Sam Altman—pause. The last time a CCP leader pledged to cooperate with the United States on a world-transforming technology, the PRC used US technology to strengthen censorship and surveillance domestically and then exported these technologies to aid other authoritarian regimes.
Unlike his predecessors, Xi has been explicit that any Western AI technologies or expertise shared with the PRC will be governed under the evolving framework of AI sovereignty. This means AI developed within the PRC will be used to strengthen authoritarian governance both domestically and abroad (HRW, April 8, 2021). PRC cyber sovereignty has already inflicted damage on global democracy, and emerging evidence points to the likelihood that AI sovereignty will be equally if not more damaging.
The PRC has attained “cyber sovereignty,” but it has key technological deficits that currently prevent it from achieving “AI sovereignty.” Based on the history of cyber sovereignty, the current dialogue on AI sovereignty, and AI’s future destructive capacities, policymakers must closely consider the consequences of sharing or collaborating with the PRC on AI.
No comments:
Post a Comment