Artificial Intelligence with Chinese Characteristics

Jinghan Zeng’s Artificial Intelligence with Chinese Characteristics arrived in 2022 into a Western policy conversation that had spent five years oscillating between two poles. On one side stood Kai-Fu Lee’s AI Superpowers, published in 2018, with its warning that China was on the brink of overtaking the United States on the strength of data volume, engineering talent, and a permissive regulatory culture. On the other stood a more sceptical camp, drawn from technologists and China analysts, arguing that the country’s headline AI achievements were exaggerated, dependent on Western hardware and open-source software, and vulnerable to serious export controls. Zeng, a Lancaster University political scientist whose earlier work sat at the intersection of Chinese politics and technology governance, set out to write past both. His book is a corrective rather than a manifesto.

Zeng’s interest lies in the political economy of AI inside a party-state. The phrase in the title is deliberate. Just as Deng Xiaoping’s “socialism with Chinese characteristics” was a formula for grafting market economics onto Leninist politics, “artificial intelligence with Chinese characteristics” describes a general-purpose technology adopted for its productive power and simultaneously bent to the requirements of one-party rule. The framing is analytical, not admiring, and it does the useful work of moving the conversation away from a straight capability race. What matters, on Zeng’s account, is not only whether China’s models are as good as America’s; it is what those models are being asked to do, who is asking, and how the asking reshapes both the Chinese state and the firms doing the building.

The book’s central thesis follows from that frame. Chinese AI, in Zeng’s telling, is not primarily a Sputnik moment for the West. It is a domestic modernization project whose international ambitions are real but secondary, and whose defining tensions are internal. The Party wants innovation but it also wants control. It wants private companies to succeed but not to become independent centres of power. It wants Chinese firms to build the surveillance and censorship infrastructure the regime relies on, and it also wants those same firms to compete for global market share against companies operating in jurisdictions that would never sign the domestic contracts. Every account that treats these tensions as either resolved or fatal, Zeng argues, misreads the situation. They are the terrain on which the industry actually operates, and they explain a great deal about why particular initiatives succeed, stall, or produce outcomes that surprise observers on both sides.

Structurally, the book moves outward from strategy to implementation to consequence. Zeng opens with the policy architecture, the sequence of central documents through which Beijing has articulated its AI ambitions. The 2017 New Generation Artificial Intelligence Development Plan is the anchoring text, with its three-stage timetable to catch the world leaders by 2020, achieve major breakthroughs by 2025, and become the primary global centre by 2030. Zeng treats it seriously without treating it literally. He is careful to distinguish declared ambition from implemented policy, and to trace how central signalling filters through ministries, provincial governments, and municipal AI parks in ways that often produce duplication, capital misallocation, and the kind of Potemkin projects that any observer of Chinese industrial policy learns to expect. The plan matters because it mobilised resources and legitimised local initiative, not because it functioned as a blueprint whose milestones were then dutifully hit.

From strategy the book turns to the firms. Zeng’s account of the Chinese AI industry is neither the hagiography that some Western venture-capital literature produces nor the caricature of a bureaucratically directed sector. The national champions — Baidu, Alibaba, Tencent, and later ByteDance — are treated as genuine research organisations with real capabilities in search, cloud, computer vision, and natural language processing. The AI specialists that Western commentary tends to lump together — SenseTime, Megvii, Yitu, CloudWalk, iFlytek — are given their distinct histories and specialisms, from face recognition to speech to public-security integration. Zeng is at his most useful when he tracks how these firms grew: the mixture of state contracts, academic incubation, foreign venture capital, and export ambition that shaped each one, and the moments when regulatory intervention or a shift in political priorities changed their trajectory. The 2020 halt of the Ant Group IPO, and the broader “rectification” of the platform economy that followed, sits in the background of the later chapters as evidence that the Party’s willingness to discipline its own tech giants is not rhetorical.

Surveillance receives its own extended treatment, because it is where Chinese AI has produced the most visible and disturbing operational outputs. Zeng walks through the layered infrastructure — Skynet, Sharp Eyes, the Integrated Joint Operations Platform in Xinjiang — and describes how facial recognition, gait analysis, voice printing, and predictive policing tools have been assembled into a system that watches particular populations with an intensity that has few historical parallels. He is careful to separate what the system claims to do from what it demonstrably does, and to note that many of its most publicised capabilities function unevenly in practice. But he refuses the technologist’s move of dismissing an unreliable surveillance system as harmless. A system that produces false positives can still terrorise; a database of biometrics collected under coercion is still a permanent record of the coercion. The Xinjiang chapter is where the book’s political stakes become impossible to sidestep, and Zeng writes it plainly, drawing on the leaked China Cables and on the field research of scholars like Darren Byler and Adrian Zenz without letting the citations substitute for his own argument.

The military dimension is treated with a similar mix of seriousness and caution. Zeng surveys the People’s Liberation Army’s public discussion of “intelligentized warfare,” the doctrinal successor to the earlier concept of informatised warfare, and traces how the concept has been elaborated in Academy of Military Science writings and in the work of PLA-affiliated researchers. He examines the machinery of Military-Civil Fusion, the effort to knock down the walls between commercial AI research and defence applications, and he notes the ways in which that policy has been more successful as an aspiration than as a set of institutional arrangements. American defence-technology analysts such as Elsa Kania have mapped this territory in more granular detail, and Zeng draws on their work while placing it inside his larger argument about how the Chinese state actually functions. The picture that emerges is of a military that takes AI seriously, that has meaningful programmes in autonomous systems, target recognition, and command-and-control decision support, and that is nonetheless constrained by exactly the same shortages of top-tier semiconductors, foreign talent, and independent evaluation cultures that constrain the civilian sector. The gap between what the PLA writes about intelligentized warfare and what it fields is real, but so is the direction of travel.

International dimensions come next. Zeng examines the export of Chinese AI-enabled surveillance systems to governments across Central Asia, Africa, and parts of Latin America, and he places that export in the wider context of Huawei’s Safe City deployments and the Digital Silk Road. He is properly cautious about the causal claims Western think-tanks sometimes make — that installing Chinese cameras causes democratic backsliding — while acknowledging that the availability of turnkey authoritarian tooling changes the calculus for governments already inclined that way. He also examines the traffic in the other direction: Chinese AI firms’ dependence on Nvidia chips and TSMC fabrication, on PyTorch and TensorFlow, on the open scientific literature, and on the researchers trained in American and British universities. The chapter on the US-China technology contest sits at the intersection of these flows and is one of the book’s soberer sections. Zeng was writing before the October 2022 chip controls tightened decisively, but he already saw the direction of American policy and understood that Beijing’s response would be a serious, expensive, and slow attempt at semiconductor self-sufficiency whose success was very much in doubt.

The book closes with governance. This is where Zeng returns to the tension that animates the whole project. China has issued more ethics guidelines and regulatory drafts on AI than almost any other government — from the Beijing AI Principles of 2019 to the Cyberspace Administration’s rules on algorithmic recommendation and, later, on generative content. Zeng reads these documents as genuine responses to genuine problems, not as pure window-dressing. Chinese consumers were tired of algorithmic price discrimination; Chinese officials were nervous about the political disruption that unregulated recommendation systems could produce; Chinese ministries wanted a share of the regulatory prestige that European counterparts were accruing through the AI Act. He also reads them as consistent with the Party’s abiding priority: any technology powerful enough to reshape how citizens see the world is a technology the Party intends to control. The ethics discourse in China, on this reading, is not a Western import waiting to bloom into liberal governance. It is a distinctive tradition being articulated in real time by a regime that has both the will and the tools to enforce it in ways that liberal democracies structurally cannot.

The reception of the book in the field has been substantial without being uncritical. Reviewers in the China-studies community have welcomed Zeng’s insistence on treating Chinese AI as a domestic political project rather than as a piece on the geopolitical chessboard, and they have found his handling of the surveillance and Xinjiang material more measured than the alternatives. Defence-technology specialists have pushed back on some of the military chapters, arguing that Zeng underweights the pace of PLA experimentation with unmanned systems, particularly the growing catalogue of loitering munitions and swarming demonstrators visible at Zhuhai and other trade shows since 2022. Others have suggested that the book’s caution about Chinese AI capabilities looks slightly different in the wake of DeepSeek’s late-2024 releases and the broader wave of competitive Chinese large language models, which have complicated the picture of a semiconductor-constrained sector permanently one generation behind. Zeng’s frame — that the interesting story is about how the Party bends the technology, not only about whether the technology works — has held up better than any specific capability assessment. In a field where books date fast, that is its most durable contribution.

For a reader building a shelf on AI and warfare, Zeng’s book sits in a particular niche. It is the political economy volume, the one to read alongside the more technical accounts to understand why particular Chinese systems get built, who pays for them, and what the state expects in return. It pairs naturally with Elsa Kania’s writing on Chinese military AI, with Jeffrey Ding’s ChinAI newsletter and his 2024 book Technology and the Rise of Great Powers, and with Kai-Fu Lee’s earlier bullish account, which it usefully corrects without discrediting. It does not replace granular reporting from the semiconductor beat, and it does not attempt the operational detail of accounts of specific weapons programmes. What it offers is a way of thinking, a set of analytical habits that make the daily news from Beijing more legible. A reader coming to the Chinese AI story cold will find the book more useful than a stack of policy briefs, precisely because it teaches how to read the briefs.

The parts of the book that are likely to age best are the analytical frame and the treatment of the party-state’s approach to platform power, both of which describe patterns that have if anything intensified since publication. The chapters most exposed to time are those on specific companies and specific capability gaps, where the pace of change since 2022 has been rapid enough to reshape particular claims. The great virtue of Zeng’s writing is that he built those chapters to be superseded. He treats capabilities as moving targets and institutions as the more durable objects of study, and that instinct is why a book about Chinese AI written before the current wave of models still repays reading in a field where most of its contemporaries do not. It is, in the end, a book about how a political system uses a technology, and the political system has not changed.