My career has been built at the edge of what is physically possible – predicting exotic quantum states, developing sensors that exploit the laws of nature, and teaching machines to reason about the physical world. For a long time, that work lived in journals and laboratories. Now, those technologies are being rapidly adopted for military applications – and not just by the United States and its allies. What just months ago seemed like a five-year problem of global technological competition has now compressed to an urgent 18-month national security sprint. The U.S. is facing a daunting race to develop and field advanced capabilities such as AI and quantum sensors on the battlefield. Many of the scientists who can help win that race are in university physics departments and at deep-tech startups. I have seen the vast potential firsthand. It must be tapped, and quickly.
The science my colleagues and I have spent our careers developing is becoming deployable at a pace that would have seemed implausible a decade ago. Quantum sensors, once confined to specialized laboratories, are shrinking, and physics-informed AI models are learning to fuse streams of sensing data in real time. The boundary between fundamental research and fielded capability is collapsing. For American national security and the military, this is both a challenge and an opening.
This is why I am concerned about the public falling-out between Anthropic and the Pentagon. The dispute – rooted in Anthropic’s insistence that its AI is not used for autonomous lethal targeting or domestic mass surveillance and the Trump administration’s decision to terminate the relationship – has left the impression of a wider chill within the deep-tech community. If Anthropic, one of the most sophisticated AI labs in the world, could not find common ground with the Department of War, why would the rest of the community want to try?
I understand that impression. But the serious scientists and technologists I know are not walking away. They are choosing to engage – because of what’s at stake. Disengagement is not principled: It is abdication. America’s adversaries are counting on technologists and the deep tech community to stay on the sidelines. They will not be so fortunate. China will not pause its next-generation computing research programs because a tech company has a principled disagreement with the Pentagon’s contracting terms. Russia will not halt its electronic warfare development. Iran will not wait.
When it comes to quantum technologies, China has kept pace with every major American advance. Every time U.S. researchers demonstrate quantum advantage, a corresponding announcement from the People’s Republic follows almost immediately – the Jiuzhang and Zuchongzhi quantum computers, for example. Their published AI models are remarkable. The capabilities we are not seeing are almost certainly even more remarkable.
On the battlefield, the gap between legacy sensing and what is now possible is equally stark. Modern platforms carry hundreds of sensor types simultaneously. Quantum sensors can exceed the limits of their classical counterparts, seeing parts of the electromagnetic spectrum where we have historically been blind. In the Strait of Hormuz, where commercial and military traffic converge, decisions must be made in seconds. The ability to fuse sensor data at the edge – rather than routing it back to analysts thousands of miles away – is the difference between early warning and catastrophic surprise. The critical challenge is no longer producing data – it is processing it. An army of analysts working at human speed cannot keep pace with what modern conflict demands. AI-driven sensor fusion can compress the time between detection and decision in ways that could prove decisive.
This is the version of AI that matters most for national security – not large language models running in data centers, but lean, physics-constrained models deployed in contested environments, with intermittent connectivity and no margin for error. Building that kind of system requires exactly the kind of expertise that the academic and deep-tech communities have developed.
As AI models are rapidly improving, our adversaries are watching the same curve. If we do not close the gap between the science being done in academic settings and the capabilities deployed to warfighters, we will have squandered a U.S. national security advantage that took generations to build. The Anthropic episode should be read as evidence of how urgently the defense ecosystem and the technology community need to build workable terms of engagement. The harder path, but the only serious one, is to stay in the room, work through the hard problems, and make sure the best technology reaches the warfighters who need it.
The best minds in physics, AI, and advanced sensing should not look away from defense. This is the moment they have been training for. I intend to show up. I hope they will too.
