The Finnish defence technology company NestAI is beginning to train its own AI models for the battlefield. The Finnish Defence Forces (Puolustusvoimat) and the Estonian Defence Forces will serve as pilot users of the first models, which are being developed for autonomy in unmanned systems and for command and control of forces.
"There is no general artificial intelligence that always works in all environments. Nor is there a single model that solves all problems. It requires continuous development," says Peter Sarlin, founder and chairman of NestAI, to the Finnish Defence Forces magazine Ruotuväki.
According to Ruotuväki, the goal is for the models to help the Finnish Defence Forces operate autonomously on the battlefield even if satellite navigation is disrupted, communications links are severed, or sensors provide incorrect information. Finland's Ministry of Defence and the Estonian Defence Forces signed a cooperation agreement with NestAI in late June covering AI development, and the Finnish Defence Forces AI Centre, together with the Estonian Defence Forces, will be the first to test the models.
In a press release, NestAI states that the first models are being trained for two tasks: battlefield autonomy, which gives unmanned systems the capability to make independent decisions, and battlefield orchestration, which supports the command and coordination of forces. The models are trained on operational data from the company's command system NestOS, supplemented by synthetic data from a digital model of the terrain and conditions along Europe's eastern flank. There, according to the company, the models can be trained on scenarios that are too rare or too dangerous to exercise in reality.
The training is carried out, according to the press release, in collaboration with partners including AMD, ELLIS Institute Finland, the Finnish Defence Forces, and the supercomputing initiative LUMI AI Factory. NestAI was founded in 2025 and has grown to 200 researchers and engineers.