The Swedish company developed the agents, their training environments and supporting tools in-house. It said that ownership gives it control over how the software behaves, how it is verified and how new capabilities are introduced.
“We have not only built our own AI agents, but we have also flown them,” said Peter Nilsson, head of Advanced Programs at Saab’s Aeronautics business area. “This demonstrates our capability to develop AI agents that can be integrated into operational fighter aircraft.”
The agents train through Saab’s AI Fighter Pilot Academy, accumulating millions of simulated flight hours across different operational scenarios. Their actions are repeatedly assessed against defined objectives, allowing them to explore tactics and improve decision-making before flight testing.
Saab said beyond-visual-range combat had now been validated under real flight conditions. Gripen E’s software-defined architecture allows new AI software to be integrated and flown on an operational fighter without requiring a dedicated experimental aircraft.
That approach shortens development cycles, while the academy’s reusable training infrastructure allows new missions, scenarios and platforms to be added without rebuilding the framework. Saab is continuing to expand the agents’ capabilities and fly them in a wider range of situations.
“The future of air combat will not be defined by individual platforms acting alone, but by networks of crewed and uncrewed systems working together as a coordinated force,” Nilsson said. “Through our ownership of the entire technology stack, we are building the capabilities needed for that future.”
The company said safety and control are incorporated from the start of development. Its work is also supported by Sferical AI, a Swedish sovereign AI initiative providing secure, large-scale computing capacity, in which Saab is a founding partner.
