Shooting the Suns

I

On September 30, 1984, Beijing Municipality opened to the public its first modern subway track, Line 2. For about US$0.035, Beijingers could now conveniently circle the capital along the route of its former imperial walls—a far cry from the impractical Line 1, which had been built by, and for a time reserved for, the People’s Liberation Army as part of Mao-era dual-use “civilian-defense” to deter a Soviet attack. The new stations were not only modern and functional; several were also adorned with large, colorful murals celebrating the great achievements of Chinese civilization.

None captured the ethos of Deng Xiaoping’s new Reform and Opening program more clearly than those installed at Jianguomen Station, “the Gate of the Founding of the Nation.” On one side of the platform, a 60-meter-long, three-meter-high ceramic mural titled “The History of Chinese Astronomy” illustrated two ancient Chinese myths, “Nüwa Creates Humanity” and “Hou Yi Shoots the Suns.” The opposite mural, made of 8,000 porcelain bricks, depicted the “Four Great Inventions” of China: papermaking, movable-type printing, the compass, and gunpowder. With their stylized lines, harmonious composition, and luminous pastels—turquoise, lilac, soft golds, and pinks—the murals offered a reprieve from the hard-edged, muscular idiom of revolutionary propaganda art that had prevailed until then. It also heralded a different kind of future.

1984 was a turning point for China. Mao had promised in 1957 to “overtake Britain and catch up with the United States.” Instead, he dragged the country into a series of catastrophic political campaigns that cost tens of millions of lives, elevating ideology over science and repeatedly subjecting intellectuals and scientists to persecution—policies that culminated in the Cultural Revolution and the nationwide closure of schools and universities from 1966 to 1976.

Upon ascending to power in 1978, Deng Xiaoping reinstated science and technology not merely as one pillar of his Four Modernizations program alongside agriculture, industry, and defense but as the core driver of the other three. Less than a month after the unveiling of the murals at Jianguomen, the Third Plenum of the Chinese Communist Party’s Central Committee adopted “the Decision on the Reform of the Economic System,” which embraced the market economy and private entrepreneurship and put China on the path of three decades of explosive “socialist market” growth. Deng’s insight—that technology was a “productive force” on its own—ultimately catapulted the country from its backward state to its status today as the world’s second-largest economy, lifting living standards for hundreds of millions of citizens and building a formidable military in the process.

The year 1984 is also notable for two far more modest events, whose significance only appears in retrospect. In April, a group of pioneering entrepreneurs led by Wan Runnan, then an engineer at the Chinese Academy of Sciences, founded the country’s first private computer company, Stone Emerging Industries. The venture quickly overtook previous state-led efforts to develop software able to handle the input of Chinese characters, and it launched a computer revolution by gaining government ministries and large state-owned enterprises as its first clients. Around the same time, a small laboratory working in the novel field of pattern recognition was established within the academy’s automation department. It is now widely regarded as having helped lay the foundations of the technology that defines our age today: artificial intelligence.

II

In their account of political and economic development, the Nobel Prize–winning economists Daron Acemoglu and James Robinson argue that successful modernization depends on a delicate equilibrium between a strong state and a counterbalancing civil society. Too weak a state and political instability cripples economic activity. Too powerless a society and the state grows increasingly predatory, hobbling entrepreneurship and discouraging innovation. Only nations that have historically managed to remain within what Acemoglu and Robinson call “the narrow corridor” have succeeded in joining the ranks of developed economies.

Where does China fit? By most measures, it is a strong state with a reasonably competent bureaucracy, one that broadly followed the path laid out by the Asian Tigers—Japan, South Korea, Singapore, and Taiwan—from the 1960s onward: industrialization through a methodical climb up the value chain of an export-oriented economy. China invested heavily in education, built vast infrastructure, channeled long-term capital into strategic industries, and used market reforms to unshackle domestic consumption. It established an extensive legal and regulatory architecture, opened its economy widely enough to attract foreign capital and technology, yet it did so selectively enough to prevent foreign firms from becoming dominant. In fact, so distinctive is the Chinese state’s capacity—rooted in a centralized, partly meritocratic bureaucracy whose history reaches back to the Western Zhou dynasty (1046–771 BCE)—that the notion other countries might replicate the “Chinese model” appears fanciful.

China’s embrace of science and technology as the fastest route to development has fostered a distinctive national confidence in progress. In the span of just four decades, from the late 1970s to the early 2020s, the country expanded the size of its economy roughly one hundredfold and per capita output nearly fiftyfold. Its technological achievements range from operating a space station and sending rovers to Mars and the far side of the moon to building 40,000 km of high-speed rail in under sixteen years. For most citizens, progress is tangible: improvements in housing, healthcare, education, transportation, connectivity, commerce, and employment opportunities.

One would be hard pressed, on the other hand, to find in Chinese society the counterforce that Acemoglu and Robinson identify as necessary to keep a nation within that narrow corridor. While Chinese citizens’ personal freedoms expanded immensely as spheres outside Party-state control grew with the private economy, the authorities have meticulously maintained strict controls on the media, academia, expression, and online content, consistently drawing the line at any form of independent organizing. The Chinese Constitution prohibits any contestation of the “Four Cardinal Principles”—the socialist road, the people’s democratic dictatorship, the leadership of the Communist Party, and Marxism-Leninism—ruling out any political challenge to the Party’s monopoly on power.

What emerges from this history is a technological trajectory shaped not by universal forces but by China’s specific inheritance: the role science and technology have played in rescuing the country from foreign aggression and internal collapse, and the institutional imperatives that flow from an authoritarian state’s need to maintain control over a vast, rapidly modernizing society. This inheritance—and the current US pressure intent on constraining it—profoundly shapes how both China’s state and society engage with artificial intelligence. That relationship is markedly different from the one in the US and it is obscured by the omnipresent framing about an AI “race.”

III

Perhaps no concept better captures the Party’s understanding of its role than the notion of shehui guanli,or “managing society.” It rests on the view that society is inherently unstable and must be regulated, monitored, and at times suppressed to avert luàn, or “chaos.” The collapse of the Soviet Union and the Chinese Communist Party’s near-death experience in 1989 imparted Chinese leaders with a lasting lesson: Social grievances had to be detected early, before they crystallized into political challenges. The Party’s response was to become the first “Darwinian-Leninist” political organization in history: a top-down autocratic system in a process of continuous adaptation to changing social circumstances.

Contrary to much Western commentary, everyday politics in China isn’t really a struggle between single-party rule and liberal aspirations—dealing with that is the work of China’s vast security apparatus. For Party and government cadres, everyday politics consists of identifying which social forces might become destabilizing and deciding whether to accommodate them, channel them into state-controlled forms, or stamp them out, all while sustaining economic development as the ultimate social pacifier. Central to this successful cycle of adaptation, power consolidation, and national strengthening are familiar tools of Chinese statecraft: allowing intellectuals to shape policy through internal state channels and harnessing technology for both economic modernization and the expansion of state power.

Few figures embody this juncture more vividly than the late-nineteenth-century reformer Kang Youwei. Kang witnessed firsthand the Qing imperial court’s inability to prevent China from being carved up by increasingly aggressive imperial powers, first European and then Japanese. He himself took part in the short-lived Hundred Days’ Reform and was forced into exile after the Empress Dowager Cixi carried out a palace coup and put a price on his head. The question of how to “save the nation” (jiuguo) had long been the major inquiry confronting Chinese intellectuals, and it became even more central as late Qing rule decayed into paralysis in the face of predatory foreign powers.

While many searched for the causes of China’s decline in its cultural or civilizational foundations, Kang advanced a different idea. His extensive travels through Asia, Europe, and the US had convinced him that in an age of intensifying competition among nation-states, a country’s strength rested on a single factor: material power. “Why Europe is strong and China is weak has nothing to do with philosophy,” he wrote, rebutting his contemporaries.

In Saving the Nation Through Material Means, his canonical work, Kang proposed a simple remedy to China’s woes: The state should send Chinese students abroad to learn from more technologically advanced countries and upon their return they should apply and pass on this new knowledge to industry, machinery, infrastructure, and military capacity—in particular the navy, whose weakness had exposed China to repeated defeats and humiliation. Everything else would follow. Embracing modern science and technology was the real solution to China’s poverty and weakness, as that would help it achieve modernization without causing the kind of social chaos that critics of reform claimed would be inevitable.

A century later, Deng Xiaoping’s prescriptions would turn out to be no different.

IV

Technology breakthroughs tend to happen like bankruptcies do: gradually and then suddenly. In December 2024, a previously little-known Chinese company startled the world with the release of DeepSeek-V3, a large language model whose performance was reported to be comparable to that of leading American systems. Its appeal rested on three features. Its smaller size meant it could be run locally wherever users were. Its training costs were much cheaper. And it was an “open weights” model, enabling developers to inspect and build upon its architecture.

Viewed from China, it represented something greater still. Starting in October 2022, the Biden administration had imposed increasingly restrictive bans on the export of advanced semiconductors to China in an explicit effort to slow its AI progress. “We need to keep as large a lead as possible,” National Security Adviser Jake Sullivan had declared. Coming a year after the latest controls were imposed, DeepSeek suggested that China had succeeded in compensating for constrained international access to hardware with homegrown ingenuity and hard work. “While the US remains obsessed with technological containment, China has shown that open innovation cannot be stopped,” a popular Chinese post declared after DeepSeek’s release. The system’s rise, the post added, was not only “a milestone of Chinese innovation” but also a “demon-revealing mirror” held up to American monopoly and hegemony. In Chinese folklore, such mirrors expose disguised spirits by revealing their true form. DeepSeek was now being cast in much the same role.

This was not the first time a Chinese technological leap was spurred by the perception that the US was pressing ahead unchecked in the development of military capabilities. Since the end of World War II the US has been constraining China’s ability to project influence through its alliances and military presence, from South Korea and Japan to Taiwan and the Philippines, forming a loose arc of containment that limits China’s ability to project power into Asia and the Pacific. Still—even though military modernization had been part of Deng’s Four Modernizations program—economic development had taken precedence over overhauling the People’s Liberation Army (PLA).

The 1991 US invasion of Iraq suddenly exposed the gap with American capabilities. As one Chinese military expert later put it, “from military theories… to the weapons and equipment… we realized [the PLA] was decades behind the Americans.” Chinese leaders responded by ordering the development of “modern regional warfare capabilities under hi-tech conditions.” The 1991 shock about US capabilities was only the first. In 2001, a US EP-3 spy plane forced to land on Hainan Island yielded data that, according to the NSA, “provided a comprehensive overview” of how the US exploited adversaries’ signal environments—and revealed China’s vulnerability. Then, in 2013, Edward Snowden’s disclosures revealed the global scope of US surveillance. Together, these episodes accelerated China’s push for technological autonomy, including the development of the BeiDou satellite navigation system and expanded cyber and information-warfare capabilities.

As important a priority as military modernization has been the harnessing of technology for the purpose of social control. In this domain, the Party-state has arguably been just as successful, especially in view of the enormous scale and speed of the social transformations China has undergone over the past half-century. From the outset, the aim was not merely to enhance the state’s repressive capacity but to provide it with greater visibility into what was happening in society.

Blanketing the country with CCTV cameras during the 1990s was the first visible deployment of this strategy, allowing, at least in theory, faster police interventions in cases of public-order disruption such as street protests. Having quickly identified at about the same time the destabilizing potential that the internet posed to the regime’s traditional control over information, media, and independent organizing, the authorities ensured that a digital infrastructure and a censorship architecture would be developed in tandem.

The next step was to establish, throughout the late 1990s and the early 2000s, the Great Firewall as a filter to keep out content, sources, and digital platforms that ran afoul of the regime’s censors. Foreign technology companies, eager to sell hardware, software, and services, helped China build this architecture of censorship before foreign digital platforms were either blocked outright, such as Twitter and YouTube, or made to comply with it. To further dilute the appeal of foreign digital platforms, the authorities encouraged Chinese companies to develop domestic copycat alternatives, which benefited from the absence of already established, powerful competitors.

Admittedly, determined Chinese users have always been able to poke holes in the system, and Chinese social media continue to host countless pockets of dissenting expression. The censorship apparatus regularly fails to prevent bursts of national uproar over social realities the state would prefer to obscure. Miscarriages of justice, corruption, oppressive patriarchal norms, food-safety scandals, medical malpractice, or the diffuse ennui of a younger generation have all, at times, triggered nationwide outcries.

But the regime’s objective has never been total censorship—in part because that would have excessively constrained the measure of openness that is indispensable to China’s efforts to catch up scientifically and technologically. Its aim has always been to prevent such moments from sedimenting into a durable civic memory of resistance.

V

Multi-layer censorship, however, is only part of the story. Much more fundamental is the close alignment between the Chinese state’s long-standing governing imperatives and the possibilities created by new technologies. Working ceaselessly to detect social demands, sources of discontent, and any developments that might damage the Party’s legitimacy, the state has found in computerized surveillance and large-scale data aggregation a providential window into society’s inner workings: how citizens react in real time to news and crises, which issues trigger moral outrage or apathy, how and where grievances spread, who amplifies them, and where and when they might warrant countermeasures.

As commercial platforms have collected ever-larger volumes of personal data, the Chinese state has improved its capacity to obtain extensive individual-level information: location and movements, social networks, financial activity, consumption patterns, behavioral data, and political views, including preferences users may never articulate publicly. For individuals placed on the security apparatus’s roster of “target populations,” the cross-referencing of surveillance records with commercial and behavioral data produces an exceptionally fine-grained form of control.

What embracing new technologies has done for the Chinese state goes well beyond reinforcing existing authoritarian tools. It has transformed the very nature of political control. Whereas the regime once relied primarily on post hoc repression and censorship, it has now shifted toward a sophisticated architecture of always-on monitoring: a persistent, granular mapping of society in real time. This is not lost on those living in Xi Jinping’s China. “I just read George Orwell’s 1984,” a friend told me recently. “Compared to us, it is really xiaowu jian dawu—a minor sorcerer meeting a master!” Digital technologies have not merely expanded the reach of state power; they have advanced the long-standing ambition of the Chinese state to render society permanently visible and intelligible.

This ambition is not new. Early forms of collective accountability can be traced to the Shang Dynasty (c.1600–1046 BCE), were systematized under the Northern Song (960–1127 CE) through the baojia system, which organized households into units for purposes of taxation and control, and were revived under the Qing (1636–1912)—with uneven results. Upon coming to power, the Communist Party drew on this repertoire. Residents’ and Village Committees, the hukou system, and the danwei tied individuals to place and work, while personal files aggregated biographical and political data.

Market reforms have since weakened these mechanisms—but under Xi they have been recombined within a “grid management” system (wangge) that divides neighborhoods into small blocks of housing monitored through CCTV networks, policing databases, and residents’ committees. Minxin Pei, a political scientist, has argued that China is now a “sentinel state” that has moved repression “upstream”: It doesn’t simply respond to political threats once they materialize but anticipates or detects and then neutralizes them before they can even coalesce.

Nowhere are the dangers of this architecture more fully realized than in the Xinjiang Uyghur Autonomous Region. A colonial setting par excellence, Xinjiang has witnessed some of the most repressive policies in China, with repeated campaigns of detention and persecution accompanied by a systematic effort to erase the cultural and religious distinctiveness of the region’s Turkic populations, perceived as potential vehicles for separatist aspirations.

The most sweeping of these campaigns began in 2016. By some estimates, up to ten percent of the 11-million strong non-Han population has been detained in “vocational legal education facilities”—in practice, detention camps—where detainees are graded for ideological “dangerousness” and assigned to detention, the formal judicial system, or surveillance. The camps have also enabled the systematic collection of personal and biometric data for a region-wide surveillance regime. As James Millward and Dahlia Peterson noted in a 2020 report based on leaked official documents, Xinjiang fused conventional policing with facial-recognition videotaping, cellphone tracking, and the mass harvesting of personal and biometric information. After feeding the data into an algorithmic system designed to identify “extremism,” individuals are sorted for imprisonment, indoctrination, or surveillance.

All of this predates the turn to artificial intelligence, which now promises to magnify the Chinese state’s already unmatched techno-surveillance capacity.

VI

For English speakers, “artificial” often evokes simulation (“artificial intelligence”), unnaturalness (“artificial flowers”), or even inauthenticity (“an artificial smile”). In Chinese, the word for “artificial” is formed by combining the characters for “person” (ren) and “work” (gong), emphasizing something made by human effort rather than set in opposition to nature. “Man-made intelligence” has a different resonance from “artificial intelligence,” if only by firmly situating AI within the world built by human labor and treating it as continuous with other technologies rather than as radically distinct from them.

The often-repeated trope that the US is more focused on general intelligence while China is more focused on diffusion and application is reflected in the use of relevant terms in both places. In English, “artificial general intelligence” (AGI) carries a sweeping, abstract ambition; the corresponding term in Chinese, tongyong, suggests simply something widely spread for use. The English terms are also widely used in China, but the distinctions track the contrast between an American fixation on AGI and a Chinese emphasis on immediate diffusion of AI throughout the economy and society.

Among American elites, no conversation about artificial intelligence is complete without some reference to its indispensability to American primacy. Eric Schmidt, Google’s former CEO, has become a leading advocate of AI as an instrument of US national power. Anthropic calls for “strong export controls to prevent authoritarian nations from developing frontier AI capabilities.” It is an understatement to say that the White House has been receptive. Both the Biden and Trump administrations have unlocked massive federal subsidies for the AI sector. Last year, invoking the need to ensure that “AI companies remain free to innovate,” Trump issued an executive order aimed at pre-empting state-level AI regulation. “America is the country that started the AI race,” he proclaimed, “and as president of the United States, I’m here to declare that America is going to win it.”

Chinese elites face markedly different constraints than their American peers. Despite China’s striking rise to become the world’s second-largest economy, it remains a developing country and it continues to portray itself as such. Around 300 million Chinese live below the World Bank’s vulnerability threshold of $8.30 per day. The country’s GDP per capita is roughly one-sixth that of the US, and China has reached the income range historically associated with the “middle-income trap,” where rapid catch-up growth can give way to prolonged stagnation. And while the US claims that China is nearing parity in AI, the reality is that China remains constrained by smaller capital markets, a poorer consumer base, restricted access to lucrative Western markets, a limited ability to attract foreign talent, far less computing capacity, and exclusion from the most advanced semiconductors and other frontier technologies.

In addition, there is a far sharper divide between those who serve the state and those in the private sphere. Government officials and Party members are said to be tizhinei, “within the system”; private entrepreneurs and the general public, on the other hand, are tizhiwai, “outside the system.” The two worlds are walled-off. Every official’s career is closely tied to measurable performance indicators designed to transmit policy objectives down the chain of command (though, as in any bureaucratic system, patronage, data manipulation, and corruption provide avenues of circumvention). Business leaders enjoy much greater autonomy by virtue of being outside the system, yet they must contend with powerful state authorities to secure licenses, credit, regulatory clarity, and access to policymakers. The relationship is one of mutual dependence. Officials rely on firms for growth and employment in their jurisdiction; companies depend on officials to operate in an intensely competitive mixed economy.

That dependence can also be a source of vulnerability: Firms may cultivate quiet ties to officials in order to shape or survive policy changes, but such ties can turn toxic during anti-corruption drives or sudden political shifts. The 2020 suspension of Alibaba’s Ant Group IPO made the point starkly. Jack Ma’s subsequent five-year eclipse from public view was a reminder that in China’s business environment, prominence offers little protection against political miscalculation, selective enforcement, or elite score-settling, any of which can bring not only financial ruin but detention, asset seizure, and severe consequences for family members, associates, and subordinates.

In short, through co-optation, coercion, and the embrace of technology, the Chinese Party-state has so far managed to stay within Acemoglu and Robinson’s developmental “narrow corridor,” all the while successfully repressing any civil-society effort that might challenge its rule.

VII

In We Have Never Been Modern, the late French thinker Bruno Latour identified a sleight of hand at the heart of modern political life: the claim that political decisions are guided by objective scientific facts that stand outside politics, values, and power. Governments justify policy choices by invoking what “science tells us,” while dissent is cast as obscurantism. After all, is it not the canonical lesson of the Enlightenment that science and its cornucopia of material benefits could only truly flourish once it had shaken off the yoke of religion and politics?

When states or tech companies invoke the language of scientific objectivity or technological inevitability—“the science is clear”; “the data show”— these phrases launder inherently political decisions about who gets what and how into the seemingly neutral register of technical necessity. But science does not, by its nature, dictate policy nor any next social arrangements. The sleight of hand that claims political choices are dictated by science operates differently in Washington and Beijing but it operates in both places. In the US, the language of inevitability is market-inflected and imperial: AI will transform everything, the race must be won, regulation is a luxury that only the losing side can afford. In China, the language is developmental and civilizational: AI is a productive force, its diffusion is a historical necessity, its deployment by the state is an extension of the Party’s mandate to make the country secure and steer society toward its flourishing.

In much Western discourse, political responsibility for the social consequences of AI is quietly evacuated under the banners of historical inevitability, technical complexity, and geopolitical necessity—what the German sociologist Ulrich Beck famously described as the “organized irresponsibility” of risk management. In China, by contrast, the Party’s absolute primacy over society mechanically assigns it the ownership of such outcomes. The core premise of China’s political system is that “North, South, East, West, and Center—the Party leads everything,” as Xi has put it. In the Chinese political imaginary, AI is not an uncontrollable force but an instrument whose effects on society the Party is understood to ultimately own.

In a June 2025 testimony before the US Congress, Jack Clark, a co-founder of Anthropic, argued that “AI built in authoritarian nations will—no matter what the personal preferences are of the people building it—be inescapably intertwined with authoritarianism.” He is right. But the inverse does not hold: AI built outside of authoritarian systems can also produce authoritarian effects. It confers on any state astonishing—almost irresistible—capacities for surveillance and social control. Consider US law enforcement’s facial-recognition dragnets or the Department of Government Efficiency’s mining of federal databases or official requests to tech companies that they provide the details of people criticizing Immigration and Customs Enforcement online.

With AI, the surveillance state and surveillance capitalism are no longer merely compatible; they are rapidly converging and reinforcing one another. The data that platforms extract from users to sell advertising is the same data that states can requisition to manage populations. The algorithmic systems that optimize engagement are the same systems that can optimize compliance. The biometric tools that platforms deploy to authenticate users or prevent fraud are the same tools that states can use to track and identify individuals across public space. The difference between China and the US increasingly is one of proportion rather than kind. Both systems fuse surveillance for monetarization with surveillance for regime security, and their differences are quickly narrowing. If AI raises a genuine political challenge, it lies precisely here, not in a debate about who might be winning a putative race.

In the Jianguomen Station mural, Hou Yi confronts a civilizational crisis when ten suns rise at once and scorch the earth. He shoots down nine of them, to restore balance and make the world habitable again. Perhaps we, too, should shoot down the extra suns of our own creation. The race for AI dominance risks driving us all toward a point of no return, where increasingly powerful states know almost everything about us while we know very little about them.

Nicholas Bequelin is a senior fellow at Yale Law School’s Paul Tsai China Center. He previously served as Asia Regional Director at Amnesty International.