AI for National Security Operators

Kevin McAleenan, CEO of BigBear.ai, in conversation with Steve Clemons at the American Growth Summit

On Sept. 29, 2026, BigBear.ai CEO Kevin McAleenan joined Steve Clemons, chairman and CEO of Widehall, at the American Growth Summit in Washington, D.C.

This transcript has been edited for length and clarity, and the questions have been condensed. Topic headings have been added for navigation.

The changing threat landscape

Steve Clemons: Kevin, you were acting secretary of the U.S. Department of Homeland Security and commissioner of U.S. Customs and Border Protection. As we said at the start of today’s program, the terrain we thought we knew is changing, and AI and quantum are moving at a rapid pace. Part of what we’re talking about today is American economic growth, and part of it is the threats to that growth. It’s both promise and disaster. How do we navigate it?

Kevin McAleenan: The threat landscape is diverse and changing. When I was acting secretary, we issued a strategic framework for countering terrorism and targeted violence. We made several predictions about the emergence of new asymmetric technologies and the attack vectors those would create.

We talked about the coming threat of artificial intelligence in the hands of cyber threat actors, for example. But the one that comes to mind most viscerally for me is that we had a picture in that report of a drone swarm ominously headed toward a city skyline. That wasn’t really a reality yet in 2019. It certainly is now, every day for our friends in Ukraine or in the Gulf, but also right here at home. Across the street, UFC Freedom 250 at the White House was threatened by a plot involving a drone attack that the FBI disrupted in advance.

These emerging technologies are commercially available, in the hands of threat actors and increasing in capability so rapidly that staying in front of them is the real challenge for security professionals.

Responding to the drone threat

Steve Clemons: Years ago, I had a friendly written debate in The Huffington Post with David Frum, who’s now at The Atlantic. I argued that drones were going to be a really big problem: Nonstate actors would get them, they could be miniaturized to deliver pretty lethal stuff, and we should work out a big transnational deal on managing them. David was very pro-drone. He wanted drones everywhere. Did we make a big mistake by not doing that deal?

Kevin McAleenan: He wanted them completely integrated into the airspace system? I think we could have been more in front of it nationally, for sure, and with allies. It was clear this was coming.

It’s very difficult, both technology-wise and in terms of jurisdiction and regulation, to make this integrated system work. You have the federal actors, the FAA and the Transportation Security Administration. They’ve taken huge steps recently in developing a counter-UAS capability that’s deployable. They supported the airports where FIFA travelers were arriving, securing that airspace and securing the stadiums.

I think we needed that kind of deployable interagency capability several years ago. It’s great that it’s developed now. But it’s also about investing early and often, and providing authority at multiple levels, not only the federal level. CBP can do important things at the border now to identify and counter drones, and more technology needs to be deployed to help them.

At the state and local level, we need an integrated system with a concept of operations that’s understood, where you have the authority to mitigate threats, not just identify that there’s something in the air and maybe find the operator. Building the ability to mitigate them with emerging tools is something we could have started earlier. To your point, we knew this was going to happen. It’s a very complex environment, but I agree there are probably some opportunities to move faster.

Keeping pace with AI and cybersecurity

Steve Clemons: Your knowledge of this space is so dense. Mine is not. A major global CEO told me recently that, even as companies look at cost controls, the cybersecurity investments they have to make keep growing. How do you keep up?

Kevin McAleenan: To your point, the information density is increasing so much that I have to spend a good portion of each day just keeping pace with tech developments. Part of it is: Are our products cyber-protected across the board at the level they need to be?

We provide a generative AI platform for the Department of War and the intelligence community. We’re the enterprise-level provider for NASA and the FBI. That has to be secure. That has to be integrated into the classified systems at the highest level of security.

Steve Clemons: Every year at the World Economic Forum in Davos, Cisco has a big digital screen that tracks, in real time, the malicious activity being launched across the internet: malware and other bad actors. So much is knowable, and it’s happening at such a staggering volume. That was going on for years before AI tools came into the picture. It’s a nightmarish picture of the threats out there. How do we think positively about American economic growth and the future without feeling like we’re going to be taken down tomorrow?

Getting new tools to operators

Kevin McAleenan: I don’t want to call it an arms race, but it’s a constant struggle between authorities trying to keep that infrastructure safe and threat actors trying to exploit it. The scale is absolutely a significant problem. I don’t envy my friends leading the National Security Agency or Cyber Command, thinking about what you just outlined.

But what I do think we can do is provide cutting-edge capabilities and get them in the hands of operators accomplishing difficult missions, counter-UAS missions, for example.

How about a transportation security officer trying to find a prohibited item in a carry-on bag? That can be done really well with AI and a computed tomography machine. How about a line of trucks at the border that needs to be inspected to prevent threats from coming into the country, such as narcotics and agricultural pests?

Steve Clemons: Why is fentanyl so hard to stop?

Kevin McAleenan: It can unfortunately be transported in extremely small amounts. That is the most devastating aspect of it. It can be secreted in a personal vehicle or transported across the border in a drone.

It’s a really difficult problem that I know DHS and CBP are pretty well sighted on. AI tools are part of the answer, honestly. You can take an X-ray scan of a vehicle and then use AI to do automated image analysis that finds items that are out of place.

Building capabilities at pace

Steve Clemons: A senior science official in the Trump administration told me that a lot is going right for America, but his up-at-night issue is that we don’t have enough Ph.D.s going into science. In the ecosystem you’re in, what are the subterranean issues you worry about?

Kevin McAleenan: I worry about the opposite side of that equation, actually. I think we’re building tools and capabilities at pace right now. Where I think we [the government and private sector] still need to improve is getting them into the hands of the security operators who need them to keep pace with the adversary.

There are a couple of different angles to that. One is scale. Applying a generative AI model and getting it into a classified system is one step. But how do you handle the training? How do you train an intelligence analyst to build agents that support their workflow, for example, and really take advantage of that capability? That’s hard work at the edge.

The other angle is making sure that government investments, and the speed at which government procures, are commensurate with the investments on the tech side. We’ve seen the rise of the neoprimes and a ton of private money flowing into defense tech. I think that’s an outstanding development, and I think we should invest intentionally in our security. But the government needs to keep pace, not just in funding from Congress, but in being able to acquire at pace.

Deploying AI in classified environments

Steve Clemons: If you talk to the CEO of Lockheed Martin, they have a totally closed system, a model that isn’t connected in any way to open source. I remember a former head of the NSA saying, “We have our systems completely detached.” Does that create a subpar world of knowledge, or is it better because it holds the secrets? Are the open-source frontier models worse off or better off than what you’re seeing in the cleared, secret space?

Kevin McAleenan: You’re actually highlighting, Steve, two different but related problems that I think we need to continue to work hard to solve.

One is: How do you get these tools loaded into classified networks, which have security parameters like Lockheed was talking to you about, but still make them fully functional? That’s hard to do. You need good models that don’t need to call back to the open internet.

The second piece is that, to build good tools, you need access to real-world data and real-world operational use cases. If those are classified use cases, or sensitive military or security professional use cases, it’s very hard for a startup building really advanced capabilities to get access to that information so it can build the tool that applies to that mission set.

That’s an area where I think the government could reach out and collaborate better, in secure sandboxes where startups and emerging companies can build better tools. That’s a hard problem.

Understanding the adversaries

Steve Clemons: Who are the bad actors? Can you describe for our audience the characteristics of the bad players out there? I don’t think we often describe who our typical enemy is in this space.

Kevin McAleenan: There are a couple of levels to that game, unfortunately. There are nation-state actors that are very capable and have the funding, but they also fund a layer of proxies. We see that in the conflict in the Gulf, with proxies that are also very capable and have some nation-state backing and access to technology and information.

We’ve seen the Houthis using AI tools to program the capability for a missile to find its target more effectively. That’s a big concern.

But what we talked about in 2019, which I’ve seen come to fruition, unfortunately, is the convergence and confluence between ideologically motivated threat actors at the individual or group level and people who just want to create havoc, cause harm and mount attacks.

We’ve seen recent cases where an attacker basically picked their ideology days or weeks before an attempted attack, because they just wanted to create chaos. That is enabled by the information ecosystem and the ability to communicate with people who have bad intentions.

It’s a variety of folks we need to worry about in the security space. And I shouldn’t forget the transnational criminal organizations. They’re incredibly well funded, and the cartels work very hard to exploit our defenses.

A changing security mission after 9/11

Steve Clemons: Peter Bergen, an old friend of mine, has a new book, “All the Presidents’ Wars,” about the 25 years of the global war on terror since 9/11. It’s very well researched, and it exposes a lot of bad decisions along the way, but it makes the point that, as a nation, we haven’t had another mass-casualty incident like 9/11. His book also makes clear there were lots of blind spots and lots of information about bin Laden that we missed. As you look at the vulnerabilities we’re blind to, could we see a repeat of that scale down the road?

Kevin McAleenan: That was the clarion call for me in serving after 9/11, for the creation of DHS and for so many professionals who work on this issue every day. Our focus was preventing terrorists and terrorist weapons from entering the country, from crossing a physical border so they could mount a coordinated attack like we saw on 9/11.

To Peter’s point, the defenses put in place by DHS, by the FBI and by our Department of Defense, operating internationally, have been incredibly successful in preventing that kind of thing.

That said, with the asymmetry created by these emerging technologies, AI and cyberspace, I think we have to be very concerned that you don’t necessarily have to have 19 operatives cross a border and train on flight manuals to conduct a mass-casualty event.

Steve Clemons: Keep protecting us. Kevin McAleenan, CEO of BigBear.ai, thank you very much.

Kevin McAleenan: Thank you.