Ukraine’s Military AI: Will Kyiv Get in Its Own Way?
A new study links Ukraine’s defense innovation to independent experimentation. Russian adaptation tests the limits of that explanation.
DefenseHub · Wednesday · AI & Technology · November 18, 2026

Ukraine's wartime innovation raises a practical question: how can a military with fewer resources discover and field useful technology quickly? A new paper by Konstantin Zhukov and Joshua Ammons argues that institutions help explain Ukraine's experience with military AI. Their focus is who gets to experiment, how frontline needs reach developers, and how failed ideas are discarded. The paper presents an institutional explanation of Ukraine's innovation advantage over Russia. That argument merits attention, but it does not establish a current ranking of combat effectiveness across the two militaries.
What We Know
Published in the Review of Austrian Economics on November 14, 2026, the paper starts from a basic problem: militaries lack the price and profit-and-loss signals that let markets sort good ideas from bad ones. The authors argue that Ukraine built a workaround through multiple independent centers of experimentation, which they call a polycentric system. Volunteer networks, frontline units, drone workshops, start-ups, and foreign partners identify needs, test solutions, and pass results into the next iteration. In their account, state funding and regulatory changes supported that process. The paper says its evidence reflects developments through early 2026, so its publication date should not be mistaken for a November inventory of capabilities.
The paper traces this structure to post-2014 decentralization, crowdfunded civic defense, transparent procurement, and Ukraine's mature technology sector. The authors distinguish those enabling conditions from the mechanism itself: independent actors trying competing solutions and learning from battlefield feedback. They characterize Russia's system as monocentric, with authority concentrated in a state hierarchy. Their argument is that this structure can expand production of selected systems while making experimentation from below harder. It is a comparison of institutional arrangements, not a claim that every Ukrainian procurement decision succeeds or that Russian forces cannot adapt.

A separate assessment from the U.S. Army's T2COM G2 Operational Environment Enterprise describes Russian deployment of AI-enabled loitering munitions, intelligence, surveillance and reconnaissance platforms, automated targeting tools, and electronic warfare systems. It identifies Chinese dual-use components and Iranian drone expertise as contributors. The report also says China, Iran, and North Korea are studying Russian and Ukrainian battlefield experience, including drone swarms and reconnaissance integration. This is an open-source U.S. Army assessment of adversary cooperation, not an independent measurement of each system's combat performance.
The two accounts therefore offer a useful tension. Zhukov and Ammons emphasize the institutional obstacles to Russian discovery. T2COM describes combat experimentation, foreign industrial support, and repeated adaptation. Foreign components do not prove that Russia merely copies finished designs, just as a Ukrainian start-up does not by itself demonstrate a superior military outcome. Comparing the systems requires evidence about development speed, reliability, production, and battlefield results.

Operational Context
The paper's distinction matters because a useful design can lose value as the enemy changes its countermeasures. A model trained to recognize one set of targets may need updating, while a drone must keep functioning as navigation and communications conditions change. Central coordination can help move a selected design into production, but slow decisions can delay the next improvement. The military problem is to combine repeated experimentation with sufficient production and support. Ukraine's latest organizational changes bring that tension into focus.
On November 11, 2026, Ukraine's Ministry of Defence announced a planned network of six specialized defense technology centers. Two directorates had already been established, covering unmanned systems and missile weapons and artillery. Four more centers are planned for air defense, electronic intelligence and electronic warfare, space policy and technologies, and deep strike systems. The General Staff is to define operational needs; the centers are to prepare technological and resource solutions and oversee implementation. The centers have different responsibilities: the electronic intelligence and warfare center is to test technologies and work with developers, while other centers focus on planning, standards, project oversight, or scaling promising systems. They are not described simply as funding offices outside the innovation process.
Whether that structure speeds development or adds another approval layer remains an operating question. Production scale is a separate constraint. In a November 16 interview reported by Defence24, Ukrainian Prime Minister Sergii Koretskyi forecast that Russia could launch as many as 6,500 jet-powered Shahed drones during November if the current pace continued. This is an attributed projection of launches, not a confirmed monthly total or a factory production figure. It illustrates why innovation speed and the volume of threats facing air defenses need separate measures.
Foreign partnerships also need to be examined platform by platform. Destinus and Shield AI announced in November 2025 that Hivemind autonomy software would be integrated into Ruta and Hornet systems, with joint flight demonstrations planned for 2026. Defense Express reported on November 8, 2026 that it remained unknown whether Ukraine had received Ruta missiles equipped with that software. The same report discussed Leonardo's Proteus uncrewed helicopter, a separate project associated with the Royal Navy. It did not establish a Leonardo helicopter delivery to Ukraine or combat use of Hivemind-equipped Ruta missiles.

My Read
Ukraine should protect the ability of a frontline unit to take a problem straight to a developer. A drone that worked last month may struggle after the enemy changes its jamming. The people using it need a way to get a fix tested quickly. That is where I see the value in having several teams working on the same military problem: one unsuccessful design need not hold up everyone else.
The new technology centers should help turn a promising design into equipment that units can obtain and maintain. That requires funding, testing, and orders large enough to sustain production. If the centers make developers wait for another round of permission before trying a change, they will have made the process worse. I would judge them by how quickly a useful improvement reaches the troops and how easily those troops can report what still needs fixing.
Russia gives Ukraine little room to get this wrong. The T2COM assessment describes Russian experimentation supported by foreign components and expertise. A Russian system only has to work well enough, and arrive in enough numbers, to create a serious problem for Ukrainian forces. Ukraine therefore needs to keep improving its equipment while also supplying enough of it. Success in a workshop matters when a unit can get the equipment, use it, repair it, and obtain the next version as conditions change.
The paper's warning about AI performance deserves close attention. If software misidentifies a target, the error may disappear inside a report of a successful strike. A buyer needs tests that can uncover that kind of failure before placing the next order. I would fund independent developers, but demand comparable evidence of what their systems do under jamming and against changing targets. Good procurement has to make it easy to try a new idea and difficult to keep selling a bad one.
What to Watch
Technology centers: whether Ukraine establishes the four planned centers and whether their testing and project oversight shorten delivery times while preserving access for independent developers.
Attack volume: reported launches and interceptions of jet-powered Shaheds during November, distinguishing observed counts from Koretskyi's projection and from production capacity.
Hivemind integration: confirmation of delivery, testing, or operational use of equipped Ruta systems, with evidence distinguishing the missile from its autonomy software.
Russian adaptation: documented changes to targeting software, sensors, and tactics, and evidence of their effectiveness after Ukrainian countermeasures change.
Recommended Sources
Russia could unleash 6,500 jet-powered Shaheds a month, Defence24
Destinus and Shield AI partnership announcement, company press release, November 19, 2025
Sources & Methodology
This analysis uses Zhukov and Ammons' academic paper, the Ukrainian Ministry of Defence's organizational announcement, T2COM's open-source assessment, an attributed Ukrainian official's forecast reported by Defence24, and industry disclosures and reporting on Hivemind integration. The academic argument is distinguished from official claims, company plans, and the author's assessment in My Read.
Sources were checked on November 17, 2026 for the edition prepared for November 18. The paper's stated evidence window runs through early 2026. Announced integrations, projected drone launches, and reported system deployments are not treated as verified combat-effectiveness measurements. Documentary photographs illustrate equipment and personnel; they do not establish the presence or performance of AI software.
Corrections or source clarifications can be sent through the DefenseHub contact page.
R. Planche · DefenseHub

Discussion about this post