Ignoring the Operator's Reality is a Strategic Vulnerability We Cannot Afford

Ignoring the Operator’s Reality is a Strategic Vulnerability We Cannot Afford

By Kevin McAleenan • March 24, 2026

The most important career decision I ever made was in 2001, leaving the legal profession to build new counterterrorism capabilities in the wake of 9/11. Almost 25 years later, we should be as concerned about the safety and security of the United States and our allies as we were in the aftermath of the worst terrorist attack in our nation’s history — but for different reasons.

We live in a new reality. The nature of threats — from the homeland to the edge — is morphing at a pace that outstrips traditional planning, procurement, and problem-solving structures. I refer to this as threat system asymmetry — the accelerating mismatch between the pace and complexity of modern threats, and the rigidity of the systems designed to counter them. The nations that solve this asymmetry will maintain and extend their strategic advantage. Those that don’t will lose it.

We must not wait to focus our talent and capital on this problem. For years, we speculated about how fast frontier AI models would advance. We also predicted a geopolitical reset. That speculation has now collided with reality. Artificial intelligence is being used by our adversaries in combat zones now, and it is in the hands of threat actors and transnational criminal organizations around the world.

Tackling Threat System Asymmetry

We must be unequivocal here. The United States must be peerless in developing and deploying — and countering — advanced AI. That cannot be a partisan debate. It’s a hard operational reality. I’ve watched adversaries close capability gaps in real time. The margin for complacency is gone.

But we spend far too little time listening to the operators who must make ever-faster decisions with ever more powerful tools. They range from warfighters and intelligence analysts behind the scenes, to officers protecting ports of entry, to those in the private sector protecting our supply chains and critical infrastructure. They are people who must make consequential decisions with imperfect information under immense pressure.

How fast we can apply frontier AI to their changing security missions is the central question for establishing enduring strategic advantage. But deploying off-the-shelf tools is not enough to maintain an edge. We must accelerate and get to the right destination. Velocity — speed with purpose and direction — beats speed alone.

The difference between AI that gets adopted and AI that gets abandoned usually comes down to one question: Was it built and deployed by people who understand the mission?

There are two sides to this question. Teams and leaders expected to operationalize mandates with hours’ notice, and teams and leaders determining where to invest billions in technology for our future security. We cannot allow the gap between these two worlds to grow. With every frontier model release, that gap risks becoming a strategic vulnerability.

Invest in Mission Understanding

We talk tirelessly about keeping humans in the loop when AI is deployed. We talk far too little about keeping the right humans in the loop when we’re deciding what to build in the first place. We must invest more in what’s too often absent in AI offerings — mission understanding — by prioritizing the voices of operators closest to the mission — analysts, warfighters, engineers, and agents — who know where technology can have the greatest impact.

I experienced this gap firsthand leading U.S. Customs and Border Protection. A decade ago, we could see the incredible advances in dual-use technology coming from startup hubs around the country, and we felt the urgent need to get those capabilities into the hands of our operators in the field — to enhance security and safeguard the lives of travelers, agents and officers.

But we didn’t know how to buy directly from emerging tech companies. We didn’t have mechanisms to communicate requirements, pilot and test the technology, or move it into production and deployment. So we created an interdisciplinary team, put uniformed operators with clear mission needs in the room with procurement experts and technologists, and gave them a simple structure to contract for — and iterate on — relevant capabilities. The results were immediate and profound — for one, we became one of Anduril’s first government customers through that process.

Fast forward to today, and no one needs to convince tech companies that defense and security customers matter. But too often, operator insight still comes only at the back end. Operators need to be engaged first — and throughout — using iterative cycles to build AI capabilities that actually deploy, scale, and hold up under pressure.

The question isn’t whether AI will transform national security — it will. It is whether the people building it understand the missions it needs to serve.