Building AI for the Battlefield

Jason Rathje—former founding director of the U.S. Office of Strategic Capital and now President of Public Sector at WebAI—shares his perspective on the future of AI, national security, and defense innovation.

In conversation with Ernestine Fu Mak, Rathje explains why distributed, on-device AI will be critical for military operations, especially in disconnected and contested environments where cloud access cannot be guaranteed. He discusses the difference between distributed AI and sovereign AI, the importance of data and model ownership, and why the future of AI will extend beyond centralized data centers. Rathje also shares lessons from building innovation programs inside the government. The conversation covers government procurement, partnerships with defense primes, and the challenges of bringing emerging technologies into government.

Along the way, Rathje reflects on why solving difficult problems requires collaboration between Silicon Valley and Washington.

Key Takeaways

• Centralized AI Won't Work in a Denied Environment

  Cloud-dependent architectures break down in contested battlefields—true warfighting capability requires models that live on the soldier's device and operate without any network connection.

• Distributed AI Is Sovereign by Design, Not by Policy

  By keeping data and intelligence on the device by default, distributed AI is arguably more secure and controllable than centralized systems that require governance frameworks to enforce ownership.

• Selling to Government Requires Full Organizational Alignment

  Standing up a public sector business demands company-wide commitment from day one—including a cleared subsidiary, dedicated compliance and policy hires, and early engagement with Congress.

• Defense Tech and Dual Use Are Categorically Different

  Companies that treat them as interchangeable underperform in both markets. True dual-use requires distinct governance, strategy, and management from pure defense-focused companies.