The Last Supper Is Over
SECWAR's speech at Starbase signals a new era at the Pentagon.
2026 has started with a bang: Delta Force’s spectacular snatch-and-grab raid in Venezuela, the ‘1979-in-reverse’ revolution brewing in the streets of Iran, the hunt for sanctioned shadow fleet tankers on the high seas, President Trump’s tough love for the primes, and much more.
Secretary of War Pete Hegseth’s speech at Starbase, Texas earlier this week deserves a place on that list. History will look back on that speech and the reforms it unveiled as an inflection point for the Pentagon in how it acquires new technology and uses autonomy to fight and win wars. Let’s hope we remember this as the moment the long hangover from the Last Supper ended and the table was set for the First Breakfast, with plenty of seats for new entrants, innovators, and heretical heroes. Or as Secretary Hegseth put it, the old era “created a closed innovation ecosystem dominated by just a handful of prime contractors. . . . Today that old era comes to an end.”
Statements like these and the actions that back them up prove that this administration is committed to unleashing innovation in our Armed Forces. Dismantling the paperwork-driven joint requirements process was the first step. Creating the Mission Engineering and Integration Activity to interface with industry was a second. The reforms announced this week are another giant leap in the right direction.
The big announcement at Starbase is an “AI acceleration strategy” to transform our Armed Forces into an “AI-first warfighting force across all domains, from the back offices of the Pentagon to the tactical edge on the front lines.”
Easy to say, hard to pull off. But that’s why the details are so enticing.
The AI acceleration strategy is designed to enable real-world action, at speed and scale. This isn’t another blue-ribbon panel to study the problem to infinity and beyond. It’s not a beard-stroking convention. It’s an aggressive effort to bring cutting-edge AI technology (and technologists) from the commercial sector into the military fold where it can be put to good use.
We have a lot of catching up to do. Bureaucratic kludge and red tape have hamstrung widespread AI deployment at the Pentagon. It takes forever to accredit cutting-edge models; the “ChatGPT moment” happened in late 2022, yet it took 18 months for the bureaucracy to approve OpenAI for use even in unclassified settings. Compounding the problem, the Pentagon has been trapped in a legal thicket because each model provider has a different acceptable use policy, some of which limit their use by the military. Additionally, the Pentagon’s AI infrastructure is woefully underpowered compared to the private sector. Classified compute is nowhere near where it needs to be to use AI in game-changing ways. Legacy programs have hoarded mission-critical data to protect their fiefs at the expense of the warfighter (and in flagrant violation of years-old decrees intended to maximize data sharing). We’ve been showing up to the prize fight late, underweight, and with our shoelaces tied together.
The AI acceleration strategy addresses these problems, leaning heavily on the private sector to close the gap between commercial and government AI capabilities. It sets an aggressive standard for accreditation: new models should be deployed “within 30 days of public release.” It slashes through the legal thicket by standardizing the acceptable use policy for AI models used by the military. It pushes for expanded access to compute from commercial hyperscalers, allowing our military to reap the benefits of the hundreds of billions in private investment in AI infrastructure. And it enforces the data decrees by requiring every corner of the Department of War to catalog its data and make it available on request unless a truly good reason for not sharing it exists. With data sharing mandatory and the playing field is level, we’ll see who is truly committed to openness—and who was faking it all along.
The strategy also focuses on real-world use cases for AI establishing seven “Pace-Setting Projects” designed to unlock AI’s potential and keep the department focused on tangible forward progress:
Warfighting:
Swarm Forge: Competitive mechanism to iteratively discover, test, and scale
novel ways of fighting with and against AI-enabled capabilities - combining
America’s elite Warfighting units with elite technology innovators.
Agent Network: Unleashing Al agent development and experimentation for Al
enabled battle management and decision support, from campaign planning to kill
chain execution.
Ender’s Foundry: Accelerating AI-enabled simulation capabilities - and
sim-dev and sim-ops feedback loops - to ensure we stay ahead of AI-enabled
adversaries.
Intelligence:
Open Arsenal: Accelerating the TechINT-to-capability development pipeline, turning intel into weapons in hours not years.
Project Grant: Enabling transformation of deterrence from static postures and
speculation to dynamic pressure with interpretable results.
Enterprise:
GenAI.mil: Democratizing AI experimentation and transformation across the Department by putting America’s world-leading AI models directly in the hands of our three million civilian and military personnel, at all classification levels.
Enterprise Agents: Building the playbook for rapid and secure AI agent
development and deployment to transform enterprise workflows.
Getting AI into the hands of the people so they can experiment, at all classification levels. Giving them armies of robot agents to compound their labor and learning. Using AI to speedrun mundane bureaucratic processes. This is what enterprise autonomy actually looks like, and OSD is laser focused on achieving it.
It’s hard to say where these projects will lead, of course. I suspect some will succeed, others will fail. But that’s the point. The Department of War is identifying problems and trusting its people to find the best way to solve them using novel technology. It’s embracing bottom-up innovation. It’s saying “bring on the revolution.”
What’s equally exciting is that Secretary Hegseth has clearly learned lessons from the Pentagon’s past and has structured his reforms accordingly.
The AI acceleration strategy acknowledges that “the person is the program.” Cameron Stanley, the new Chief Digital and Artificial Intelligence Officer, will oversee the strategy in conjunction with the new, single CTO of the Department of War, Emil Michael. Each of the Pace-Setting Projects will have a single lead, with full control and full accountability for outcomes. We know there are new Rickovers, Schrievers, and Boyds in the Pentagon. Now we’ll find out who they are. This is their chance to rise to the occasion.
Just as we’ve seen in other aspects of this administration’s acquisition policy, the AI strategy embraces the need for speed. The great inhibitors of speed in the Department of War are the thousands of invisible non-statutory rules, regulations, and processes inside the department itself. So the strategy establishes a “Barrier Removal Board” that will meet every month to hear from the people in the implementation trenches about what’s stopping them from succeeding. And the board is empowered to bulldoze obstacles, not just study them. Instead of creating an oversight layer to say ‘No’ to the builders, this one exists to say ‘No’ to the bureaucracy.
Secretary Hegseth is taking a big swing here. He has produced a strategy that meets the moment and recognizes the stakes of success and failure. As I noted last month, the choice we face is whether the technology of the future will advance freedom and human flourishing or tyranny. Secretary Hegseth clearly understands this stark reality:
The question before us is not whether or not the most powerful technologies of this century will reinforce free societies. Is it going to reinforce our free societies, or will that technology be shaped and twisted by malign regimes that seek to use those technologies for control and coercion?
We’ve moved beyond comprehension to action. This strategy has the ambition to overhaul the Pentagon for the AI age: a full gut renovation, not a shiny coat of paint on the old system. Now we have to execute.