US Air Force taps 5 secret firms to build robot wingman brains

Building a stealthy drone airframe is hard, but teaching it to dogfight like an ace pilot without human intervention is the real challenge. The U.S. Air Force calls this autonomous flight software the brain of its planned robot wingmen.

Now, the military is funding five different companies to build that brain. Just don't ask who they are.

Speaking at the Air Force’s Life Cycle Industry Days in Dayton, Ohio, Brig. Gen. Jason Voorheis confirmed that five vendors have received funding to develop autonomous control systems for the Collaborative Combat Aircraft (CCA) program. The Air Force is deliberately keeping the company names secret to protect their technical designs from foreign espionage.

A black-box race for the cockpit

Col. Timothy Helfrich, senior material leader for the advanced aircraft division, explained that mission autonomy acts as the brain of the vehicle, which justified a stricter classification posture. He did share one breadcrumb: the list includes a mix of traditional defense giants and non-traditional software vendors.

While the Pentagon won't name names, defense watchers point to startups like EpiSci and Shield AI as likely contenders. Both companies previously provided the AI flight software tested by Air Force Secretary Frank Kendall during his AI-piloted F-16 flight earlier this year.

The Air Force will put the five vendors through a progressive downselect process based on ongoing performance. Eventually, the pool will narrow to one or more providers whose software will power Increment 1 of the CCA program.

Splitting hardware and software

Whoever wins the software bake-off will see their code loaded onto physical aircraft built by General Atomics and Anduril. Both hardware vendors must successfully fly their prototype airframes before the Air Force makes its final production call in 2026.

Voorheis noted that the military fully intends to hold an open competition in the 2026 timeframe, and it is equally likely that both General Atomics and Anduril will make it into production. The service isn't planning a fixed 50-50 or 60-40 production split; procurement volumes will rely entirely on prototype performance and production capacity needed by U.S. forces and global allies.

Looking further ahead, the Air Force will kick off concept refinement for Increment 2 in fiscal 2025, which may involve international co-development and co-production agreements.

Why it matters

The Air Force's secretive software bake-off highlights a massive shift in military procurement: treating combat aircraft like smartphones where the operating system is completely separated from the hardware manufacturer.

Traditionally, prime contractors like Lockheed Martin or Boeing delivered tightly integrated, proprietary platforms that made third-party upgrades painfully slow and expensive. By selecting General Atomics and Anduril for the physical airframes while pitting five software developers against each other, the Pentagon prevents vendor lock-in and injects Silicon Valley-style speed into defense software.

However, this decoupled model carries major engineering risks. In split-second aerial combat, integrating third-party mission autonomy with unfamiliar flight control surfaces could introduce latency or unpredictable edge cases. The finger-pointing between airframe builders and software vendors during flight testing will be fierce.

For defense tech startups, this proves that proprietary AI algorithms can command serious Pentagon dollars without requiring a century of heavy manufacturing heritage. Future air superiority is officially being outsourced to the keyboard.