In August 2023, a deputy secretary of defense stood before a crowd at the National Defense Industrial Association’s Emerging Technologies conference and laid out a plan to flood the battlefield with cheap, expendable robots. Kathleen Hicks called it the Replicator Initiative, and she was blunt about the reason: China. Washington would not try to match Beijing ship for ship. Instead, it would overwhelm adversaries with thousands of autonomous systems that could be lost and replaced without crippling the force.
Nearly three years later, that bet is no longer theoretical. By spring 2026, the Pentagon has moved Replicator from concept to early fielding, China has published a national roadmap for humanoid robots whose capabilities map directly onto military needs, and Russia continues to experiment with autonomous strike systems on the battlefield in Ukraine. No binding international agreement governs any of it. The result is a three-way competition over lethal autonomy that is accelerating faster than diplomacy can follow.
Washington’s autonomous push
The Replicator Initiative remains the most thoroughly documented piece of this competition. Hicks’s 2023 speech named specific goals: field “all-domain attritable autonomous systems” in the thousands, spanning air, sea, and land, on a timeline measured in months rather than the Pentagon’s usual decades. By early 2024, the Defense Innovation Unit confirmed the program was on track, with initial capabilities moving toward operational units.
The logic is straightforward. China fields the world’s largest navy by hull count and has invested heavily in long-range missiles designed to keep American carriers at a distance. Rather than building more billion-dollar warships to close that gap, the Pentagon wants swarms of low-cost drones and unmanned vessels that can saturate a contested zone. Lose a hundred, and you launch a hundred more.
But the program carries unresolved questions. The Congressional Research Service has flagged concerns about how Congress will fund Replicator across multiple budget cycles, how the Pentagon will integrate autonomous platforms into existing command structures, and how much latitude these systems will have in selecting and engaging targets. The FY2026 defense budget request includes continued funding for autonomous systems, though the full scale of Replicator spending remains spread across multiple line items and is difficult to track from public documents alone.
Beijing’s robotics roadmap
China’s contribution to this competition looks different on paper but points in a similar direction. In late 2023, the Ministry of Industry and Information Technology published formal guidance for the “innovative development” of humanoid robots. The document, issued as notice No. 193, lays out a national roadmap calling for breakthroughs in autonomous sensing, real-time decision-making, dexterous manipulation, and safe human-robot interaction.
On its face, this is industrial policy. The targets are framed around manufacturing, elder care, and disaster response. But under China’s civil-military fusion doctrine, the line between commercial technology and defense capability is deliberately thin. Advances in mobile, semi-autonomous humanoid platforms could support military logistics in contested environments, hazardous engineering tasks, or eventually direct ground operations. Analysts tracking the People’s Liberation Army note that China’s concept of “intelligentized warfare” treats AI integration as a defining feature of future conflict, not an add-on.
No publicly available PLA document directly links the MIIT humanoid robotics program to a specific weapons system. That gap matters. The assumption that military adoption will follow rests on pattern recognition: China has previously channeled civilian drone and cyber technology into defense applications through the same fusion framework. The inference is reasonable, but it remains an inference, not a confirmed program of record.
Russia: battlefield experimentation, thin documentation
Russia is the hardest of the three to pin down. Moscow has used a mix of drones, loitering munitions like the Lancet series, and electronic warfare systems in Ukraine, and battlefield reporting suggests incremental moves toward greater autonomy in some of these platforms. The Marker unmanned ground vehicle has been tested in operational settings, and Russian defense officials have spoken publicly about AI-enabled warfare as a priority.
Yet the evidentiary basis is far weaker than what exists for the United States or China. Among the sources reviewed for this article, there are no primary Russian Ministry of Defense strategy papers or declassified reports that lay out an autonomous weapons development program with the specificity of Hicks’s Replicator speech or Beijing’s MIIT notice. Coverage of Russian AI ambitions leans heavily on think-tank assessments, translated media reports, and battlefield observation, all of which are useful but sit several rungs below official policy documents on the reliability ladder.
This matters for how seriously to take claims about a “three-way” race. U.S. and Chinese acceleration is documented in their own governments’ words. Russia’s role is real, shaped by urgent wartime needs in Ukraine, but its long-term autonomous weapons trajectory is harder to verify. Sanctions, industrial bottlenecks, and the grinding demands of a conventional war may be limiting how fast Moscow can move, though the fog around Russian programs makes confident assessment difficult.
The missing guardrails
What ties these three efforts together is not just the technology but the absence of any binding framework to govern it. Discussions at the United Nations Convention on Certain Conventional Weapons have produced years of debate over lethal autonomous weapons systems, or LAWS, without yielding a treaty or even an agreed definition. In 2023, the United States issued a Political Declaration on Responsible Military Use of Artificial Intelligence, gathering dozens of endorsements, but the declaration is voluntary and carries no enforcement mechanism. China has not signed it. Russia has not signed it.
Meanwhile, the competitive dynamic is self-reinforcing. Washington explicitly names China as the reason for Replicator. Beijing frames its modernization as necessary to maintain strategic balance. Each side’s public commitments give the other a reason to accelerate, and neither has paired its production timelines with binding limits on how far autonomy in targeting will go.
The systems being built today may start with a human operator in the loop, approving each engagement. But once thousands of autonomous platforms are fielded across multiple domains, the operational pressure to expand their independence will be intense. Swarms operating in communications-denied environments, for instance, may need to make targeting decisions faster than a human supervisor can review them. The technical architecture being locked in now will shape what is politically possible later.
What the next moves look like
For defense planners, the near-term picture is concrete. The Pentagon is scaling production of attritable drones and unmanned vessels, with industry partners like Anduril and Shield AI competing for contracts. China is pouring state resources into embodied AI and humanoid robotics, with provincial governments and major manufacturers aligning behind the MIIT roadmap. Russia is iterating on the battlefield, using Ukraine as a live testing ground for autonomous and semi-autonomous strike systems.
For everyone else, the practical question is whether governance can catch up to procurement. As of spring 2026, it has not. The systems now entering production were designed in a policy vacuum, and the diplomatic conversations that might constrain them are moving at a fraction of the speed. If the pattern holds, lethal autonomy will not remain an edge case in military planning. It will become a central feature of great-power competition, with rules written after the fact, if they are written at all.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
