Shield AI's Brandon Tseng on 10 years building AI pilots for fighter jets and the new developer platform launching in May

Summary

Brandon Tseng co-founded Shield AI in 2015 with his brother after serving as a US Navy SEAL across three deployments. The company's core product is what Tseng calls an AI pilot — self-driving technology for unmanned systems that operates without GPS, without communications, and without a remote human operator, including swarming and teaming capabilities.

The milestone timeline is worth knowing. Shield AI put the first AI pilot on a battlefield in 2018, deploying it on a quadcopter sent into buildings ahead of Special Operations forces in Iraq, Afghanistan, and Syria. The same technology was used by Israeli counterterrorism forces in October 2023 to rescue hostages. On the fixed-wing side, Shield AI won the DARPA AlphaDogfight trials in 2020, flew the first autonomous F-16 in 2023, ran the first AI-versus-human F-16 dogfight the same year, and in 2024 had the Secretary of the Air Force fly its AI-piloted F-16. The company was a finalist for the Collier Trophy — awarded previously to the Wright Brothers and Chuck Yeager — losing to a NASA asteroid mission.

Tseng is clear that the F-16 work was primarily to burn down technical risk for the next generation of uncrewed fighter jets, not an end product in itself. On the autonomy maturity question, he argues AI pilot performance is already at superhuman levels, and that the human involvement required to augment those systems is becoming minimal. The hardest remaining problem is not flight performance but safety certification — convincing the Air Force and Navy that a 20,000-pound aircraft operating at closure rates of 1,000 miles per hour within 50 meters of another aircraft will do exactly what it is supposed to do.

Developer platform

The commercial expansion announced alongside Shield AI's recent funding round is a developer platform going to general availability in May 2025. The framing Tseng uses is deliberate: Shield AI spent a decade building internal tools, infrastructure, and data pipelines to make its own AI pilots work, then asked how to get a million AI pilots into customer hands within ten years. The answer was to package those tools and open them to the broader defense industrial base and autonomy industry — explicitly modeled on how Amazon Web Services commercialized Amazon's internal infrastructure.

The platform's claimed efficiency gains are significant: Tseng says it enables teams to reduce engineer headcount and time-to-market by 10 to 50x. As a concrete benchmark, Shield AI is now achieving first flights on jet aircraft with two engineers in six weeks. Tseng notes that if these tools had existed in 2015, Shield AI might have accessed F-16 testing far earlier than the seven years it actually took.

M&A posture

Shield AI has built an acquisition function and is actively scanning. Tseng says a small IP acquisition will be announced within weeks. Larger deals need to clear multiple strategic filters — market access, product roadmap alignment, and meaningful needle-moving for the business — given the executive bandwidth they consume. The overall posture is opportunistic but disciplined, with a bias toward acquisitions that extend the core autonomy platform rather than adjacencies.