Air Force Swaps AI Pilots Mid-Flight
The U.S. Air Force successfully demonstrated the ability to swap AI pilots on an Anduril drone mid-mission. In the test, Shield AI’s Hivemind system took over control from a previous algorithm, a major step toward a modular, "app store" model for battlefield autonomy. This validates the concept of deploying and updating AI agents in real-time.
This demonstration occurred on Anduril's YFQ-44A, a candidate for the Air Force's Collaborative Combat Aircraft (CCA) program. The test involved taking off with one AI pilot, activating Shield AI's Hivemind system to run through test sequences, and then seamlessly switching to Anduril's own Lattice for Mission Autonomy to complete the same tests before landing. The technical backbone for this AI swap is the Autonomy Government Reference Architecture (A-GRA). This government-owned standard separates the aircraft's core flight controls from the tactical "mission autonomy" software, creating a stable platform where different AI agents can be plugged in, much like apps on a smartphone. This prevents vendor lock and allows for rapid, competitive upgrades. This capability is central to the CCA initiative, which aims to pair at least 1,000 autonomous drones with crewed fighters like the F-35 and the next-generation fighter. The goal is to create an affordable mass of aircraft for missions like strikes, reconnaissance, and electronic warfare, extending the reach and effectiveness of the piloted fleet. The Air Force plans to spend over $8.9 billion on CCA programs between fiscal years 2025 and 2029. The Anduril YFQ-44A and General Atomics' YFQ-42A are the two prototypes competing for the first increment of the CCA program. Both platforms are now flying with third-party mission autonomy software; the YFQ-42A recently flew with Collins Aerospace's Sidekick AI. The Air Force is expected to select a final winner for both the aircraft and the mission software contract in 2026. This modular approach, often referred to as Open Mission Systems (OMS), is designed to accelerate acquisition and reduce long-term costs. By using government-owned, non-proprietary standards, the DoD can foster competition and allow smaller tech companies to develop components and software for major platforms, breaking the traditional prime contractor model. Companies like Shield AI have leveraged programs like the Small Business Innovation Research (SBIR) and Strategic Funding Increase (STRATFI) to develop these advanced AI capabilities. These programs are designed to help innovative startups bridge the "valley of death" between initial funding and securing major defense contracts, feeding the exact kind of modular technology demonstrated in the YFQ-44A flight.
