In 2018, Zeng Yi — a senior executive at NORINCO, China's third-largest defense company — delivered a public speech that Western analysts initially dismissed as bluster. "In future battlegrounds, there will be no people fighting," he declared, predicting that by 2025 lethal autonomous weapons would be commonplace and that their proliferation was "inevitable." Eight years on, the PLA's progress toward that vision is no longer dismissible. It is accelerating.

China's military modernization has crossed a threshold. The People's Liberation Army is no longer merely studying autonomous weapons — it is deploying them, testing swarm tactics at scale, integrating commercial AI into weapons systems, and systematically dismantling the boundary between its civilian technology sector and its military-industrial complex. The implications for regional stability, deterrence theory, and the global arms race are profound and immediate.

$225B+ China's declared defense budget 2025 — actual military AI spending estimated 3–5x higher by independent analysts

The GJ-11 Sharp Sword: China's First Stealth Combat Drone

The Gongji-11 (GJ-11), designated "Sharp Sword," represents Beijing's most significant public demonstration of autonomous strike capability. First shown publicly at the 70th anniversary military parade in 2019, the flying-wing UCAV resembles a scaled-down B-2 Spirit and is designed for deep penetration of adversary air defense networks. With a reported range exceeding 4,000 kilometers and a payload capacity of 2,000 kilograms, the GJ-11 is engineered specifically for the kind of long-range autonomous strike missions that traditionally require human crews.

What makes the GJ-11 strategically significant isn't just its stealth profile — it's the autonomy architecture embedded within it. AVIC (Aviation Industry Corporation of China) has published research describing AI-driven terrain-following algorithms, autonomous target recognition using dual-band infrared and radar fusion, and multi-vehicle coordination protocols that allow groups of GJ-11s to operate as a coordinated strike package without continuous datalink back to a human controller. In a contested electromagnetic environment — precisely the scenario China is preparing for — this matters enormously.

A 2025 CSET analysis of Chinese defense contract records found that over 2,857 AI-related defense contracts were awarded between January 2023 and December 2024 alone, many relating to autonomous flight systems and target recognition. The GJ-11 program absorbs a significant portion of AVIC's combat UAS research budget, which defense analysts estimate has grown by over 40% annually since 2021.

Operational Doctrine: Manned-Unmanned Teaming

The PLA is not simply building autonomous weapons to replace human pilots — it is building an ecosystem in which human pilots lead unmanned wingmen. Chinese military publications from 2024–2025 describe "intelligentized air combat" formations in which a J-20 or J-35 serves as the "mothership," directing swarms of expendable UCAVs toward high-value targets. The GJ-11 and its derivatives would fill the loyal-wingman role, carrying ordnance into heavily defended airspace while the crewed aircraft remains outside the threat envelope.

WZ-7 Soaring Dragon: Intelligence, Surveillance, and the Kill Chain

While the GJ-11 dominates discussions of Chinese strike autonomy, the WZ-7 Soaring Dragon (Xianglong) HALE drone is arguably more immediately consequential for regional security. Operating at altitudes above 18,000 meters with an endurance exceeding 10 hours, the WZ-7 is China's answer to the RQ-4 Global Hawk — a persistent ISR platform capable of monitoring the entire first and second island chains simultaneously.

The WZ-7's significance lies in its role as a kill-chain enabler. Chinese naval doctrine increasingly relies on what PLA strategists call the "Anti-Access/Area Denial Reconnaissance-Strike Complex" — a layered system in which persistent aerial ISR, space-based sensors, over-the-horizon radar, and AI-cueing systems feed targeting data to hypersonic missiles, submarine-launched weapons, and surface ships. The WZ-7, flying 24/7 over the South and East China Seas, is the optical and signals intelligence node that makes this kill chain function.

⚠ THREAT ASSESSMENT: ISR FUSION

PLA doctrine documents captured in open-source translations describe a real-time AI fusion layer — designated internally as the "Battlefield Information System" — that aggregates WZ-7 SIGINT/IMINT, satellite data from the Yaogan constellation (47+ reconnaissance satellites as of 2025), and submarine acoustic surveillance arrays. The system is designed to generate targeting-quality tracks on U.S. Carrier Strike Groups within 15 minutes of initial detection, with AI-cueing reducing the latency of targeting decisions from hours to under 10 minutes.

Drone Swarm Programs: From Concept to Battlefield Reality

China's most disruptive autonomous weapons development is not a single platform — it is the systematic development of drone swarm technology designed to overwhelm adversary point-defense systems through mass, coordination, and autonomy. Since 2015, when the PLA began investing heavily in swarming research, Chinese teams have broken multiple world records for coordinated UAV swarms. In 2017, CASC demonstrated a 1,000-drone swarm over Guangzhou. By 2022, indoor swarm demonstrations involved over 3,000 coordinated platforms.

But demonstration swarms performing choreographed light shows are categorically different from combat swarms operating in contested electromagnetic environments. The PLA has been working to bridge exactly that gap. A March 2025 paper from the PLA National Defense University — reviewed and translated by analysts at The Diplomat — described how Chinese researchers have developed swarm architectures built for fully autonomous execution of the kill chain in urban settings, specifically engineered to operate without communications links to defeat jamming.

The architecture, described as a "self-organizing intelligent swarm network," allows individual nodes to sense, decide, and act locally using onboard AI while maintaining emergent coordination through proximity-based communication protocols that function even when GPS and RF links are severed. This is not future research — Chinese procurement records show mass production contracts for subsystems matching this description were issued in 2023 and 2024.

4,000+ Estimated daily FPV/combat drone production capacity across Chinese defense manufacturers as of Q4 2025

Commercial Drone Sector as Military Force Multiplier

DJI controls roughly 70% of the global consumer drone market. Autel Robotics holds another significant share. Both are Chinese companies. Both have contributed — sometimes directly, sometimes through dual-use technology transfer — to the PLA's drone warfare capabilities. When Ukraine and Russia both began converting off-the-shelf DJI Mavics into grenade-droppers in 2022, they were inadvertently demonstrating something the PLA already understood: that China's commercial dominance in drone manufacturing translates directly into military-industrial leverage.

The PLA's civil-military fusion strategy (军民融合, jūnmín rónghé) formalizes this leverage. Under Xi Jinping's directive, civilian technology companies are legally required to support national defense needs upon request. This means that DJI's component supply chains, manufacturing capacity, and technical expertise are theoretically mobilizable for military production — a capability that has no Western equivalent at comparable scale.

DeepSeek Integration: AI That Costs $6M to Train, Deployed in Weapons

January 2025's most significant geopolitical event may not have been a military deployment or a treaty signing — it was the release of DeepSeek R1, an open-weight AI model trained for approximately $6 million that matched the performance of systems costing hundreds of millions to develop. For the PLA, the implications were immediately recognized.

A Reuters investigation published in October 2025, reviewing hundreds of research papers, patents, and procurement records, found that the PLA had rapidly integrated DeepSeek AI into military weapons systems ranging from autonomous combat vehicles to drone swarms. The efficiency of DeepSeek's architecture — achieving high reasoning performance at a fraction of the compute cost of American counterparts — makes it particularly suited to embedded deployment in autonomous weapons systems where processing power is constrained by size, weight, and power (SWaP) limitations.

Chinese researchers published papers in 2025 describing DeepSeek-derived models running onboard small UAS platforms for autonomous target classification, terrain navigation, and cooperative swarm behavior. The cost efficiency matters: if a country can deploy capable battlefield AI for $1,000 per drone rather than $100,000, the number of autonomous weapons it can field increases by a factor of 100. That arithmetic fundamentally changes deterrence calculus.

Naval Autonomous Vessels: The JARI and the USV Fleet

The PLA Navy (PLAN) is building what may become the world's largest fleet of autonomous surface vessels. The Joint Autonomous Research Institute (JARI), established by the PLAN in 2020, has overseen development of multiple classes of unmanned surface vehicles (USVs) designed for roles including mine countermeasures, anti-submarine warfare, and — most provocatively — anti-ship missile delivery.

The JARI Marine Lizard USV, a 15-meter catamaran capable of 42 knots and armed with surface-to-surface missiles, was publicly revealed in 2018 but has since been significantly upgraded. By 2025, China's Wuhan University of Technology and CSSC (China State Shipbuilding Corporation) had published designs for next-generation USVs with ranges exceeding 1,500 nautical miles, AI-driven obstacle avoidance, and the ability to operate in coordinated swarms of up to 100 vessels. In a Taiwan Strait scenario, such a capability could overwhelm Coast Guard and naval forces through sheer numerical saturation.

The PLAN has also demonstrated interest in autonomous underwater vehicles (AUVs) for both surveillance and mine-laying. The HSU-001, China's first publicly disclosed large AUV, was paraded in 2019. Since then, Chinese naval publications have described AUV systems with endurances of 30+ days, AI-driven acoustic pattern recognition for submarine tracking, and modular payloads that could include both sensors and munitions.

⚠ STRATEGIC RISK: TAIWAN CONTINGENCY

RAND Corporation modeling from 2025 estimates that in a Taiwan Strait conflict, a coordinated PLA autonomous systems attack — combining USV swarms, UAS saturation strikes, and hypersonic missile salvos — would generate target-engagement demands exceeding the capacity of current U.S. and Taiwanese defensive systems by a ratio of approximately 8:1. The mathematics of autonomous attrition favor the attacker when production costs drop below a threshold where quantity overwhelms quality-based defenses.

AI-Guided Missiles and Hypersonic Precision

China's missile programs have integrated AI at multiple points in the kill chain. The DF-21D and DF-26 "carrier killer" anti-ship ballistic missiles incorporate terminal guidance systems that rely on AI-driven image recognition to identify and track moving naval targets during the terminal phase — a computational problem that would be impossible to solve with pre-programmed guidance alone, given that aircraft carriers move and can execute evasive maneuvers during a missile's 20-minute flight time.

The YJ-91 anti-radiation missile and its descendants incorporate cognitive electronic warfare capabilities — onboard AI that can identify and track emitting radar systems even when those systems are frequency-hopping or using low-probability-of-intercept waveforms. This effectively means the missile hunts radar systems autonomously, without requiring updated pre-mission targeting data from its launch platform.

Chinese researchers have also published extensively on AI for hypersonic maneuvering reentry vehicles (MaRVs). The DF-17 hypersonic glide vehicle, traveling at Mach 5–10 in its terminal phase, faces an extreme control challenge: the aerodynamic forces at hypersonic speeds require microsecond actuator adjustments that exceed human or traditional algorithmic control capabilities. AI-based trajectory optimization, trained on thousands of simulated flight profiles, is the enabling technology that makes precision hypersonic weapons possible.

Civil-Military Fusion: How China Erased the Boundary

China's Military-Civil Fusion Development Strategy (MCFDS), formally adopted as national strategy in 2017, is the legal and institutional framework through which Beijing extracts military value from civilian innovation. Unlike the U.S. system, in which defense primes and commercial tech companies operate in largely separate ecosystems (with Defense Department programs attempting to bridge the gap), China has designed a system of legal compulsion, financial incentives, and organizational integration that makes the separation structurally impossible.

Key mechanisms include: the National Intelligence Law (2017), which requires Chinese organizations and citizens to "support, assist, and cooperate with national intelligence efforts"; the Military-Civil Fusion Development Commission, chaired by Xi Jinping himself; and a network of "dual-use technology research institutes" embedded within civilian universities but funded by the CMC Science and Technology Commission.

A CSET Georgetown analysis published in September 2025, reviewing 2,857 AI defense contracts from January 2023 to December 2024, found that while legacy state-owned defense enterprises (NORINCO, AVIC, CASC) remain dominant contractors, private technology companies — including AI startups, commercial drone manufacturers, and semiconductor firms — are receiving an increasing share of PLA AI contracts. The trend suggests civil-military fusion is working as designed: civilian AI innovation is being systematically harvested for military application.

2,857 AI-related PLA defense contracts awarded in 24 months (Jan 2023–Dec 2024) — CSET Georgetown analysis

PLA 2027 Modernization Target: What It Actually Means

Xi Jinping's 2020 directive to "accelerate drone warfare research and training" came alongside a broader mandate for the PLA to achieve "high-level mechanization, informatization, and intelligentization" by 2027 — the centenary of the PLA's founding. Western analysts have frequently focused on the 2027 date as it relates to Taiwan invasion planning, but the intelligentization component of that goal is independently significant.

"Intelligentized warfare" (智能化战争, zhìnénghuà zhànzhēng) is not merely a buzzword in PLA doctrine — it represents a coherent theory of how autonomous systems, AI-driven decision support, and networked sensors change the nature of military operations. The doctrine envisions conflicts decided not by the force that can field more soldiers or heavier tanks, but by the force that can generate, process, and act on battlefield information faster. In this framework, autonomous weapons systems are not merely useful tools — they are the primary mechanism by which "cognitive dominance" (制脑权) is achieved.

By 2025, multiple indicators suggest China is on track for meaningful — if not complete — achievement of its intelligentization goals. The PLAN has fielded multiple classes of autonomous surface and underwater vehicles. The PLAAF has integrated UCAVs into training exercises with manned aircraft. The PLA Ground Force has conducted battalion-level exercises involving robot dogs for reconnaissance, autonomous resupply vehicles, and AI-assisted fire control systems. And the Strategic Support Force, responsible for cyberwarfare, space operations, and electronic warfare, has AI integrated throughout its operational systems.

The Global Proliferation Risk

China's autonomous weapons development poses a secondary risk beyond direct military confrontation: proliferation. China has already exported armed drones to at least 16 countries, including the Wing Loong II (exported to UAE, Saudi Arabia, Pakistan, Egypt, and others) and the CH-4 Rainbow (Iraq, Jordan, Saudi Arabia, Nigeria). As Chinese autonomous weapons mature and production costs fall, the export market will expand to include genuinely autonomous lethal systems — not merely remotely piloted ones.

Unlike nuclear weapons, autonomous weapons have no dedicated nonproliferation regime. There is no Autonomous Weapons Non-Proliferation Treaty, no export control framework specifically targeting lethal autonomy, and no international inspection regime. When China begins exporting GJ-11-class autonomous UCAVs — a development most analysts expect within five years — the global proliferation landscape will shift in ways that current defense planning frameworks are not equipped to address.

// KEY INTELLIGENCE POINTS
  • China's PLA has awarded 2,857+ AI defense contracts in 24 months — autonomous systems are production reality, not R&D aspiration
  • DeepSeek military integration enables capable onboard AI at fraction of U.S. compute cost — asymmetric quantity advantage
  • GJ-11 stealth UCAV designed for autonomous deep strikes; WZ-7 feeds real-time ISR data to hypersonic kill chains
  • Naval USV and AUV programs target carrier strike group saturation — RAND estimates 8:1 target generation advantage in strait scenarios
  • Civil-Military Fusion law makes Chinese commercial tech sector legally mobilizable for defense — DJI dominance has military implications
  • PLA 2027 "intelligentization" target is on track across PLAAF, PLAN, and Ground Force exercise evidence
  • Zero autonomous weapons export control regime exists — proliferation of lethal autonomous systems to 16+ current Chinese drone-import nations is structurally inevitable

Assessment: The Gap Is Narrowing, and May Already Be Closed in Some Domains

U.S. defense planning has long assumed a comfortable technology lead over China in advanced military systems. That assumption is now contested in the autonomous weapons domain and may be wrong in specific areas. China has matched or exceeded U.S. capabilities in several relevant dimensions: drone swarm scale, naval USV development, hypersonic missile production rates, and — through DeepSeek — the cost efficiency of battlefield AI deployment.

The areas where the United States retains meaningful advantages — satellite communications infrastructure, fifth-generation fighter capability, undersea warfare, and the ability to project power globally from carrier strike groups — are precisely the areas that China's autonomous weapons programs are designed to target and degrade. The GJ-11 targets air superiority. The DF-21D targets carriers. The USV swarms target surface combatants. The autonomous submarine systems target undersea dominance. This is not coincidence — it is deliberate design.

Western policymakers who frame the autonomous weapons competition primarily in terms of ethical debates about "killer robots" are missing the strategic reality. China is not debating whether to deploy lethal autonomous weapons. It is determining how to deploy them most effectively. The only relevant Western response is to make comparable choices — faster, with better systems, and with a coherent doctrine for how autonomy integrates into existing force structures. Time is not on Washington's side.