TL;DR
Ukraine’s Defense Ministry is using Avengers Labs to give approved defense companies access to annotated combat drone footage inside a protected Brave1 Dataroom. The arrangement lets companies train models on real battlefield imagery while Ukraine keeps the finished AI, a model-for-data exchange that could shape drone warfare and allied defense technology.
Ukraine’s Defense Ministry has turned Avengers Labs into a controlled access point for approved defense companies to train AI models on annotated combat drone data inside the Brave1 Dataroom, a model-for-data exchange that gives Kyiv the improved systems while keeping raw battlefield footage under state control.
The program, described in Ukrainian government reporting and by Thorsten Meyer AI’s summary of the initiative, gives domestic and foreign developers a way to train computer-vision systems on visual and thermal footage gathered during tens of thousands of drone sorties. The material is said to include camouflaged vehicles, targets seen at night or in poor weather, multiple sensor views, and drones operating under electronic warfare pressure.
Companies do not receive the underlying footage, according to the source material. Training, validation and fine-tuning take place inside the Brave1 Dataroom, a protected environment built with the Ministry of Defense, the Ministry of Digital Transformation, the Armed Forces, military-intelligence researchers and Palantir. More than 100 Ukrainian companies are said to have access, with allied developers joining through Avengers Labs.
The arrangement is unusual because the main payment is not cash. Firms gain access to scarce battlefield data, and Ukraine keeps the improved models they produce. The Defense Ministry’s operational Avengers platform is already used to detect, classify and track hostile targets from drone and camera feeds; ministry reporting cited by the source material says it flags about 12,000 enemy units a week and feeds the VEZHA streaming module in DELTA, Ukraine’s battlefield-management system.
Avengers Labs
Ukraine’s Ministry of Defense is renting access to the world’s only large-scale, real-war computer-vision dataset. The terms: train your model inside the protected Dataroom — Ukraine keeps the finished AI.
Inside the Dataroom
- Structured visual & thermal imagery of aerial and ground targets
- Hard cases: camouflaged armor, night, fog, rain, multiple sensors
- Feeds the Avengers platform inside the DELTA / VEZHA system
- Focus track: automatic detection & interception of enemy drones
The goal
- 100% of frontline drones with onboard machine vision
- Autonomous navigation in GPS-denied / jammed (EW) skies
- Autonomous Shahed interception — human keeps the trigger
- Scaling vs. Shahed launches rising ~35% / month
Combat Data Becomes Leverage
For AI developers, the data is the scarce asset. Computer-vision systems for battlefield drones can fail when training sets do not match combat: smoke, camouflage, low light, dirt, fog and electronic attack. Ukraine's claim is that its front-line data captures those conditions at scale.
For Kyiv, the model-for-data exchange turns wartime observations into a defense-industrial asset. It could speed better target detection, reduce operator workload and support low-cost interceptor drones against Shahed-type attacks, while keeping sensitive footage under state control. For allies, it offers a testing ground for systems they may later buy or adapt, but it also raises policy questions about data governance, export controls and human control over targeting.
AI training dataset for drone surveillance
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Drones To Dataroom
Brave1 was launched in 2023 as Ukraine's defense-innovation cluster, bringing government, military users, investors and developers into a faster wartime procurement pipeline. Mykhailo Fedorov, previously associated with Ukraine's Army of Drones effort, became defense minister in January 2026, according to the source material.
The Dataroom work follows Ukraine's wider push to use AI and cheap interceptors against growing Russian missile and drone attacks. The Washington Post reported in January 2026 that Fedorov signed an agreement with Palantir to build an AI Dataroom using four years of sensor data and imagery to train systems for strike prediction and autonomous interceptors. Avengers Labs appears to extend that logic from a state data project into a controlled partnership program for companies.
"We have a clear plan about how to stop Russia in our skies."
— Defense Minister Mykhailo Fedorov, quoted by The Washington Post in January 2026
Autel Robotics EVO II Dual 640T V3, 640x512 Thermal Imaging Drone, Remote ID Supported, 0.8'' CMOS&50MP 8K RYYB Sensor,10+ Temp Measurement, D-RI Ranges, Thermal Analysis Tool, for Search&Rescue/Power Inspection/Public Safety/Hunting
🏆【AUTEL ROBOTICS USA AUTHORIZED】We are the longest and largest distributors of Autel Robotics. Over the years, we have...
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Model Performance Still Unproven
Several points remain unconfirmed from public material. The exact legal terms for foreign participants, how Ukraine vets companies, how much data each partner can use, and whether models can be exported after training have not been detailed. The reported figure of millions of annotated frames and tens of thousands of sorties comes from Ukrainian officials and source summaries; independent auditors have not publicly verified the dataset's size or quality.
It is also unclear how far Avengers-based detection can be pushed toward autonomous interception. Ukrainian officials describe human oversight for lethal decisions, but public accounts do not provide a full technical picture of how target IDs, operator authorization and error handling work under battlefield pressure.
NVIDIA Jetson Orin Nano Super Developer Kit
The NVIDIA Jetson Orin Nano Developer Kit sets a new standard for creating entry-level AI-powered robots, smart drones,...
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Battlefield Tests Will Decide
Avengers Labs' next milestone is performance in active use: whether partner-trained models improve detection rates, reduce false positives and work on drones facing jamming, weather and limited onboard computing. Ukraine has set a goal of putting machine vision on all front-line drones and advancing Shahed interception with human-controlled firing decisions.
Readers should watch for new Ministry of Defense reporting on weekly detections, interceptor trials, foreign partner participation and any rules governing how models trained in the Dataroom can be used outside Ukraine.
DJI Mini 4K Camera Drone Combo, Drone with 4K UHD Camera for Adults, Under 249 g, 3-Axis Gimbal Stabilization, 10km Video Transmission, Auto Return, 2 Batteries for 62-Min Max Flight Time, QuickShots
No Registration Needed - Under 249 g, this drone with camera for adults 4K does not require FAA...
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Avengers Labs?
Avengers Labs is the partnership platform tied to Ukraine's Defense Ministry and Brave1 that lets approved defense companies train AI models on annotated combat drone data inside a protected Dataroom.
What data are companies allowed to use?
The source material says the Dataroom includes structured visual and thermal imagery from real drone missions, including aerial and ground targets seen in difficult combat conditions such as night, fog, rain, camouflage and electronic warfare.
Do companies receive raw combat footage?
No, according to the described terms. Companies work inside the Dataroom, and the raw data stays there. Ukraine receives the improved models produced from the training process.
Is Ukraine building fully autonomous weapons?
Public descriptions focus on detection, navigation and drone interception support. The source material says a human keeps control of the firing decision, but the full technical and legal setup has not been made public.
Why would foreign firms accept these terms?
Real annotated battlefield data is hard to obtain, especially for drones operating under jamming, weather and live fire. Access to that data could help firms build models that perform better in conditions ordinary test ranges cannot reproduce.
Source: Thorsten Meyer AI
