1 hour ago
1
“The [Defense] Department (DoD) is requesting a massive increase for DAWG. For those in the audience that may not know, DAWG is the Defense Autonomous Warfare Group [tasked with rapidly developing, testing, and fielding large numbers of un-crewed systems and drones] and it's going from the $225 million [in fiscal year 2026] up to the $55 billion for fiscal year 2027. And at the same time, we're integrating the AI-driven [Artificial Intelligence-driven] targeting with those autonomous munitions at a pace that DoD directive 3000.09 was not designed to contemplate.”
That was Senate Armed Services Subcommittee Chairman Sen. Joni Ernst (R-Iowa) on May 19, speaking during a hearing on the science and technology priorities contained in the Fiscal Year 2027 Defense Authorization Bill and the Future Years Defense Program.
The 85-minute subcommittee session covered not only the proposed sharp budget increase in new autonomous weaponry, but also the race that’s going on between the U.S., China and other countries to integrate AI into offensive and defensive warfare. Ernst was questioning Defense Undersecretary for Research and Engineering Emil Michael when she brought up DoD directive 3000.09 which, as updated in 2023, established policy “for developing and using autonomous and semi-autonomous functions in weapon systems, including armed platforms that are remotely operated or operated by onboard personnel.”
Ernst asked: “Secretary Michael, has the department formally reviewed whether the current governance framework is actually keeping pace with DAWG’s growth and then how do we overcome that?”
Michael responded, “It absolutely needs updating…because of the threat environment -- what's possible by the adversary -- and partly because of the lessons we learned in Iran.” He explained that the U.S. wants “autonomous mine-seeking capabilities” for the Hormuz Strait, and the Trump anti-missile Golden Dome “has an autonomous element to it, a space-based interceptor that could …hopefully get a Chinese hypersonic missile in the first 90 seconds of launch before it separates into decoys and multiple munitions. So there are going to be different risk levels with autonomous and we have to account for them in our policies. My belief is that will change more frequently than it has in the past than it ought to, to be consistent with our values, consistent with the threat environment, and consistent with the technology development.”
In his opening remarks, Michael described concerns with China, when it comes to the AI competition.
“From a national security standpoint, this is another case of our adversary, the main adversary, China, you know, taking our IP [intellectual property] from our American development labs that have spent hundreds of billions of dollars [on AI] by the end of the next couple of years…And they're “distilling” those [AI] models, which means effectively copying them for a fraction of the
price, taking off the guard rails for them, which means they could be used in ways that they're not intended to be used, which is very dangerous for us, whether it's cyber as a cyber weapon, as a biological weapon, as a chemical weapon.”
“So the threat is real,” Michael said, adding, “We have to stay ahead on chips, power, innovation and capital formation and that gives us this six to 12 month lead and maybe we could extend it. In the last Commerce [Department] NIST (National Institute of Standards and Technology) evaluation, our lead had increased by a few months against the Chinese.”
Discussion of AI in directed energy/laser weaponry was one area that caught my eye.
“Directed energy is one of my top critical technology areas,” Michael said, “So it is a focus by us. The science for directed energy is largely done and now we're in the engineering phase of it. So the engineering part of it makes it cheaper, smaller and more proliferated. We now have a suite of directed energy products that go from low-end to high-end and now we have to scale production of those. The things that are helping are Golden Dome [anti-missile defense systems], because they have a big reliance on directed energy…And because the commitment was made to the President [Trump] that we're going to have a demonstration that includes directed energy in our Golden Dome architecture, there's a lot of energy going into that.
Michael added, “While we're going to have multiple demonstrations, the primary demonstration where it [laser technology] demonstrates a lot of capabilities will be summer of [20]28.”
When Subcommittee Chairman Ernst asked Undersecretary Michael, “What are we doing to ensure that the transition pathway from that [AI weapons] prototype to actual production is actually functioning,” he gave as an example Castelion, a company he said, “developing low-cost hypersonics less than half-a-million-dollars per missile relative to the $50 million per missile we pay today.”
Backing up his statement, I found that last April 24, the U.S. Navy announced it had awarded Castelion a $105 million to continue efforts to integrate its Blackbeard hypersonic strike weapon onto the F/A-18 fighter/bomber and transition the system to an Early Operational Capability in 2027 for carrier-based operations.
And on May 13, DoD announced “once Castelion achieves testing and validation, the Department will award a two-year multi-year procurement contract for a minimum of 500 Blackbeard missiles annually, with options to extend for up to five years. To further encourage Castelion's self-funded facility expansion, the Department is actively seeking the necessary authorizations and appropriations to purchase over 12,000 Blackbeard missiles over five years.”
Subcommittee Ranking Minority Member Sen. Elissa Slotkin (D-Mich.) raised two questions that took up a good part of the panel’s time.
“Given the strategic importance of winning, I cannot for the life of me understand two decisions that have been made,” Slotkin said. “Number one [was] the decision to sell Nvidia chips to the
Chinese, giving them not our most sophisticated, but some of our most sophisticated chips and chips they do not have.”
“Secondly,” she said, “I do not understand picking a fight with one of the few [AI] companies, Anthropic, that's in all of your [DoD] systems. All of you [the military services] use Anthropic right now, to the point where we've named them a supply chain risk, and all of you are supposed to be divesting from Anthropic in the next two months.”
“On the chips question,” Michael said, “this is a debate within the technology industry which is if you sell an adversary older chips, do you slow down their domestic production of equivalent chips because they become reliant on your technology?…If they become used to the American stack, is that net better for the American AI proliferation? And that's a debate.”
Michael added, “And the White House has decided that if we gave them two versions behind chips that we'd be able to preserve our dominance on the programming language, and make it less encouraging for them to develop their own domestic chip industry to catch up.”
As for the withdrawal from Anthropic, Michael said, “What we're worried about with the terms of service that they [Anthropic] had, and their posture toward the department [DoD], which when they questioned the [Venezuela President] Maduro raid, and whether their software was used inappropriately [in his kidnapping], gave us the sense that this was not a reliable partner to deal with…in conjunction with their written terms of service which prevent the use cases that we would like to advance into -- battlefield management, directing interceptions, developing weapons systems.”
Michael explained, “Google, who's been a longtime partner of the [Defense[ Department, Microsoft, Nvidia, real big companies with proper corporate governance, went through their legal teams and agreed to our terms of all lawful use cases, where Anthropic would not. So that should say something that our terms weren't unreasonable.”
However, last week news stories reported that White House Chief of Staff Susie Wiles personally overruled the Pentagon's supply chain risk designation for Anthropic when it came to the company’s contract with the National Security Agency (NSA), which collects and processes electronic foreign intelligence communications.
The revised Anthropic contract with NSA drops the previously contested "any lawful use" Pentagon language, and adds an explicit clause restricting use of Anthropic tools for processing data on American citizens.
In this case, the White House appears to have supplanted the Pentagon in setting the rules for AI contracts. It remains to be seen how these conflicting decisions will be worked out.
The Cipher Brief is committed to publishing a range of perspectives on national security issues submitted by deeply experienced national security professionals. Opinions expressed are those of the author and do not represent the views or opinions of The Cipher Brief.
Have a perspective to share based on your experience in the national security field? Send it to Editor@thecipherbrief.com for publication consideration.
Read more expert-driven national security insights, perspective and analysis in The Cipher Brief
English (US) · 