AI Drones operating in the Ukraine War
AITech

AI Drones operating in the Ukraine War

By 4 February 2026No Comments

Eric Schmidt’s Secret Entry into Modern Warfare

Google ex-CEO Eric Schmidt has developed AI drones that are being used in live, real-world combat deployments in Ukraine.

Once launched, they are no longer controlled by any user, but make decisions internally; capable of completing targeting and strike missions completely autonomously. This represents a significant shift in how lethal force can be applied on the battlefield, moving into the realm of sci-fi films and raising serious questions about accountability and the future of warfare.

Eric self-funded most of the project initially, selling hundreds of test units to Ukraine at cost price in order to obtain real-world feedback on their performance. Today, hundreds of thousands of these drones are being shipped to Ukraine, mostly paid for through indirect funding from the EU.

Background on Eric Schmidt

Eric Sschmidt

Eric Schmidt led Google from 2001 to 2011 and later chaired the US National Security Commission on Artificial Intelligence, where he argued that AI would be decisive in future conflicts. Following Russia’s full-scale invasion of Ukraine in 2022, Schmidt became directly involved in efforts to apply commercial AI technologies to Ukrainian defence needs.

By mid-2022, Schmidt was travelling to Ukraine, meeting with military officials and advocating for low-cost, software-driven drones as a way to counter Russia’s numerical and industrial advantages. These efforts evolved into a concrete project in August 2023 with the formation of Swift Beat Holdings, operating through an Estonian entity, Volya Robotics OÜ.

The company’s stated aim is to produce inexpensive, mass-manufacturable drones using onboard AI to function in contested environments where GPS and communications are degraded or unavailable.

The start-up business kept a low profile

In early 2024, the project adopted the name “White Stork,” referencing Ukraine’s national bird and signalling its close alignment with Ukrainian defence efforts. Later that year, the project was renamed again to “Project Eagle” to avoid confusion with an unrelated humanitarian organisation operating under the same name.

Throughout this period, the company maintained a low public profile. It operated out of Eric Schmidt’s family office, Hillspire, in Menlo Park, and avoided traditional defence-industry publicity. Recruitment focused on experienced engineers from companies such as Google, Apple, SpaceX, and US government agencies, as well as staff connected to Schmidt’s philanthropic organisations.

Within its first year the team had grown to more than a dozen staff and was working closely with Ukraine’s Ministry of Defence, incorporating battlefield feedback directly into design and software updates.

Operational use in Ukraine

By 2025, Project Eagle’s drones were reportedly flying regular combat missions against Russian forces. Ukrainian facilities began producing the drones locally, with output measured in the hundreds per month. The systems are supplied at cost and are designed to be disposable, favouring quantity and adaptability over longevity.

The drones fall broadly into two categories: one-way attack drones and interceptor drones. Unit costs are reported to be in the order of a few hundred US dollars, making them far cheaper than conventional missiles and well suited to swarm or saturation tactics.

Testing initially occurred in the United States before shifting to Ukraine for operational trials. Feedback from Ukrainian operators—particularly around electronic warfare and GPS jamming—led to rapid software iteration. By late 2025, the systems were also informing NATO and European counter-drone training efforts, including exercises involving US, Polish, and Romanian forces.

About the Technology

These drones have onboard computer vision and machine-learning models to identify, track, and pursue targets using live camera feeds rather than external guidance. This allows them to function in areas where GPS and communications have been jammed. They are deliberately built from accessible commercial off-the-shelf (COTS) components, such as Raspberry Pi-class microcomputers, to enable massive scale, low production costs (around $400–$500 per unit for attack versions), and rapid iteration. Project Eagle aims for production targets exceeding 6,000 units per month, primarily for models like Bumblebee and Hornet.

Drone Models:

  • Bumblebee: A one-way attack drone designed to visually identify and pursue targets such as vehicles or personnel after launch. It features a small quadcopter (multicopter) form factor with an 11 lb (~5 kg) high-explosive payload, cruise speed of 62 mph (~100 km/h), and operational range exceeding 90 miles (~145 km). Sensors include two airborne electro-optical/infrared (EO/IR) cameras for real-time visual recognition of soldiers, bunkers, vehicles, or other drones. Navigation relies on Visual Inertial Odometry (VIO) and optical flow for fully GPS-denied environments, with redundant communications (including internet/broadband control from hundreds of km away). Reports indicate autonomous hit rates exceeding 70 percent once a target has been designated by a human operator, who locks on and toggles autonomy mode—allowing the drone to chase, adjust speed/heading, and strike independently. Over 1,000 combat flights were logged by spring 2025, targeting logistics, armor, and bunkers.
  • Hornet: A fixed-wing strike variant of Bumblebee, sharing similar AI and autonomy features. It has a wingspan of approximately 2 meters, an 11 lb (~5 kg) payload, 62 mph cruise speed, and range over 90 miles (~145 km), with EO/IR sensors for targeting.
  • Surveyor: An interceptor drone designed to disable enemy drones through collision (kinetic ram) or proximity detonation with a small warhead. It adopts a compact fixed-wing form factor with a shark-like fuselage, squashed X-rudders, and large vertical endplate rudders on the wings. Maximum speed exceeds 280 km/h (175+ mph), making it capable of intercepting fast threats like jet-powered Shahed/Geran-3 variants. Sensors include EO/IR for machine vision, plus radio frequency (RF) detection and radar signature support. It supports fully autonomous terminal homing or remote piloting, with resistance to GPS/EW jamming. Cost per interceptor is approximately $15,000, and it launches from pneumatic tubes or pickup-truck beds, with parachute recovery on misses.
  • Merops CUAS: A broader counter-unmanned aerial system intended to detect and neutralise threats such as Shahed-type loitering munitions. It integrates Surveyor interceptors with ground-based detection, offering a low-cost setup for monthly production in the hundreds. The system combines AI for detection and response, drawing on EO/IR, RF, and radar sensors for comprehensive threat neutralization.

The longer-term vision, articulated by Schmidt in public remarks, involves networked groups of autonomous drones capable of cooperative behaviour, dynamic target allocation, and resistance to jamming or spoofing. This includes “swarming” tactics where drones communicate to evade countermeasures and adapt in real-time, positioning the technology as a potential game-changer in AI-driven warfare.

Strategic and Ethical Implications

The deployment of these systems places Ukraine at the centre of a major shift in military technology. While the drones are framed as defensive tools to offset Russian aggression, they also demonstrate how quickly lethal autonomy can move from theory to battlefield reality.

The ethical concerns are substantial. Once launched, these systems can pursue and strike targets without further human input. This raises unresolved questions about responsibility in the event of misidentification, civilian harm, or unintended escalation. There are also concerns about proliferation, particularly if similar systems are copied or sold to less constrained actors.

Schmidt has spoken publicly about the need for “ethical AI” in defence, but critics argue that live deployment in an active war undermines meaningful oversight and accelerates an international arms race in autonomous weapons.

What is clear is that autonomous weapons are no longer hypothetical. They are already being used, refined, and normalised in real combat, with global consequences that are only beginning to be understood.

References:

Related Posts

AIMusic

Suno

Seb
Seb29 November 2025
AIFuture

Should I be worried about AI?

Seb
Seb12 November 2025
AIPeople

Lvmin Zhang

Seb
Seb25 April 2025

Leave a Reply