Apr 15, 2026 · 58m · big-technology
The Pentagon's AI Plan + Behind the Anthropic Fight — With Under Secretary of War Emil Michael
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this Pentagon interview, host Alex Kantrowitz speaks with Under Secretary of War Emil Michael about the integration of artificial intelligence into military targeting, drone warfare, and defense decision-support systems. Michael addresses the Pentagon's contentious supply chain dispute with Anthropic, the imperative of human command oversight, and ongoing reforms to modernize defense procurement through Silicon Valley partnerships.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 28.8% of the talking time here. How this is scored →
Alex as informed peer 5.5 Guest teaching 3.3 Guest disagreement 2.8 Alex pushing back 3.9
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 40:02 Emil rejects double standards on AI catastrophic risk
Emil forcefully dismisses the notion that the Pentagon is overreacting, turning AI lab founders' own existential warnings about bioweapons and mass unemployment back on them to defend the supply chain ban.
Hardest push from Alex ▶ 47:45 Alex quotes Judge Rita Lin on First Amendment retaliationAlex confronts Emil with Judge Rita Lin's direct finding that the Department of War illegally punished Anthropic for its public press statements.
Biggest teaching moment ▶ 34:00 Emil educates on statutory surveillance authoritiesEmil corrects external misunderstandings of Pentagon operations by detailing statutory limits under the National Security Act of 1947 and FISA, reframing Anthropic's contractual surveillance demands as legally misdirected.
Alex holds their own ▶ 37:20 Alex cites AISI benchmark data on Claude MythosAlex demonstrates superior technical preparation by quoting AISI evaluations showing Claude Mythos completing an automated 32-step cyber corporate takeover to challenge the logic of excluding frontier models.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| AI in Warfare: Autonomy Analogies and Precision Threat Discrimination | 6 | 3 | 1 | 2 | Alex demonstrates substantial technical literacy regarding Pentagon programs by detailing Cameron Stanley's demonstration of Maven Smart System and Target Workbench. Emil expounds on the human context window and visual discrimination of drone swarms collaboratively. | |
| Targeting Workflows and Digital Transformation in Defense | 6 | 2 | 1 | 2 | Alex cites reporting from Pirate Wires detailing historical reliance on Excel and PowerPoint workflows prior to recent digital transformation. Emil expands on the timeline from ChatGPT in 2022 to secure government networks. | |
| AI Architecture: Orchestration Layers vs Skynet Stereotypes | 5 | 4 | 2 | 3 | Alex presses on whether LLMs are integrated directly into targeting or the kill chain. Emil corrects Hollywood 'Skynet' misconceptions, categorizing enterprise, intelligence, and warfighting use cases. | |
| Operational Speed, Decision Friction, and Information Advantage | 5 | 4 | 3 | 5 | Alex challenges the removal of operational friction, suggesting slow legacy tools like spreadsheets might provide a necessary moral buffer. Emil rejects the premise, asserting that speed and richer data reduce casualties and maintain decision integrity. | |
| Autonomous AI Agents and Maintaining Human Judgment in Warfare | 5 | 4 | 3 | 4 | Alex questions whether military competition will inevitably tempt the Pentagon to automate end-to-end lethal decisions. Emil firmly distinguishes between defensive automated tactical routines and strict human-in-the-loop strike governance. | |
| Lessons in Drone Warfare: Attritable Systems and Cost Asymmetry | 4 | 3 | 1 | 1 | Alex highlights asymmetric air defense economics, noting the millions spent to down cheap drones. Emil elaborates on shifting procurement toward cheap, mass-attritable systems. | |
| Project Lucas and the Drone Dominance Initiative | 5 | 3 | 2 | 3 | Alex asks about specific initiatives like Project Lucas and Ukrainian partnerships. Emil clarifies the Drone Dominance initiative, supply chain divestment from foreign adversaries, and counter-UAS capabilities. | |
| AI in Cyber Warfare and Transition to the Anthropic Standoff | 4 | 2 | 1 | 1 | Alex transitions the discussion from physical drone swarms and cyber vulnerabilities to the procurement standoff with Anthropic. | |
| Contract Impasse: Surveillance Red Lines and Legal Authorities | 6 | 5 | 4 | 4 | Alex outlines Anthropic's contractual red lines regarding mass surveillance and autonomous weapons. Emil counters that Anthropic's stance was marketing theater, clarifying statutory authorities and existing ethical directives. | |
| Supply Chain Risk Classification and Claude Mythos Cyber Concerns | 8 | 3 | 4 | 7 | Alex directly confronts Emil on designating Anthropic a supply chain risk, citing the AI Security Institute's evaluation of Claude Mythos completing an end-to-end cyber range attack. Emil frames the model's offensive cyber potency as precisely why unaligned governance poses risk. | |
| Vendor Lock-in, Distillation, and Frontier Model Parity | 6 | 4 | 4 | 6 | Alex challenges Emil's claim that frontier model alternatives are interchangeable today, pushing back with 'but they don't yet.' Emil argues that rapid model distillation quickly closes performance gaps across vendors. | |
| Evaluating Existential AI Rhetoric and Supply Chain Security | 5 | 4 | 5 | 5 | Alex questions the unprecedented nature of labeling an American software provider a supply chain risk. Emil responds by referencing AI labs' own apocalyptic warnings regarding catastrophic risks and biological weapons. | |
| Model Refusal Risks, CDC Precedents, and Multi-Vendor Ecosystems | 6 | 3 | 4 | 5 | Alex questions the technical risk by noting models are hosted on AWS GovCloud without vendor kill-switches and asks why multi-model routing doesn't solve refusal risks. Emil cites a real-world CDC pathogen research refusal to illustrate operational vulnerability. | |
| Refuting Political Motivations Behind the Anthropic Decision | 6 | 3 | 4 | 5 | Alex brings up Pete Hegseth's public tweets about 'woke' companies to probe whether the Anthropic ban was politically driven. Emil denies political motives, pointing to months of technical negotiations and Anthropic's revenue growth. | |
| First Amendment Legal Battles and Historical Parallels with Google | 7 | 4 | 4 | 6 | Alex cites Judge Rita Lin's legal ruling regarding First Amendment retaliation against Anthropic's press engagement and draws historical parallels with Google's 2018 Project Maven revolt. Emil dismisses the First Amendment claim as legally unworkable in federal procurement. | |
| The Evolution of Silicon Valley Defense Culture | 5 | 2 | 2 | 4 | Alex asks Emil about reported personal financial holdings in SpaceX and xAI. Emil details his formal ethics disclosure, divestments, and recusals. | |
| Reforming Pentagon Procurement: Milestone Contracts and Defense Tech VC | 5 | 3 | 2 | 3 | Alex prompts Emil on reforming defense procurement away from legacy cost-plus contracts toward milestone-based commercial agreements, closing with a humorous exchange about the Pentagon Pizza Index. |
Statements from this episode (35)
Michael: AI enables drone threat discrimination far beyond human capacity
“You could do much more, be more precise, be more specific about what you're going after, what you're defending how you know, and the precision is really what's interesting to me because if you can use AI to detect and discriminate and, by discriminate, I mean,…”
Michael: Synthesizing multi-source data improves military human decision-making
“If you could synthesize that information so they can make that decision, and you're using more sources by definition almost, the data and choices are going to be better choices.”
Michael: AI decision support improves military operators by an order of magnitude
“The idea of taking one person and giving them the power of 10 people makes them better at what they do by an order, potentially an order of magnitude.”
Michael: Palantir serves as an AI orchestration layer for military data
“I think the genesis of what Palantir does is an orchestration layer on top of data streams that we, data that we put in it and say, here's the data we would normally use for any battlefield operation, plus an AI to help you synthesize it. So all those things a…”
Michael: Pentagon AI systems require human authorization, not an autonomous 'Skynet'
“It's not like, it's not a Skynet thing, like, no. It is a tool like any other tool that you might have on your computer, or in your war room, or with your team, except it's on your computer, visualized where, but you still have Checks and balances. You still h…”
Michael: AI targeting tools keep identical rules of engagement and approvals
“It's, you have the same rules of engagement, the same approval system. What you now have is better aggregation and synthesis of the data that you would already use to make that decision.”
Michael: Operational speed prevented U.S. casualties in Venezuela operation
“Look what happened in Venezuela. The speed at which that execution of that operation happened meant that we didn't have any casualties on our side.”
Michael: Authoritarian regimes will replace untrusted generals with AI decision-making
“I worry about other countries who don't have that, using AI to take humans out of the decision-making progress. They don't trust their generals. Because of graft, because of because they don't have the expertise, and they start to use machines in place of huma…”
Michael: China is conducting the largest military buildup in world history
“I think if you think about the biggest military buildup in world history in China, and you think of it, you've seen the purge of the generals and sort of the military hierarchy there,”
Michael: Pentagon AI agent pilots are limited to enterprise operations
“But in terms of agents, we've had some agent pilots at our enterprise level. Remember I was talking about the enterprise corporate level? Just to do the mundane things we have to do every day. But those things are not sort of where we're at the war fighting le…”
Michael: DoD Has 35-Page Directive Governing Human Oversight of AI Systems
“And we've written that there's a 35 page directive at the DoD that talks about
Human oversight and how we manage these systems, and we're constantly updating that and making sure we have the right controls on it.”
Michael: The Term 'LLM' Will Disappear as Models Become Multimodal
“So that's why LLMs, I think, is going to go away as a
The term, because they're not large language models only.
They're visual.
They're going to be used for robotics.
They're going to be used for a lot of things.”
Michael: Defending against cheap drones requires affordable mass-attritable systems
“And to protect your exquisite targets on your side against a very cheap drone, you have to use expensive countermeasures. And so the lesson there is how do you turn the dial from, maybe we should have more mass attritable, Weapons like drones or counter drones…”
Pentagon incorporates Ukrainian companies to onshore drone manufacturing and tactics
“We just launched our drone dominance program, and I think there was two Ukrainian companies in it. We're gonna, like, you know, onshore manufacturing here and take some of the learnings with them here to help us with our kind of smaller drone you know scenario…”
Pentagon Drone Dominance Program Requires Zero Supply Chain Dependency on Adversaries
“So that is one of the requirements is that the drones that we use at the Drone Dominance Program don't have a dependency on, on adversaries.”
Michael: Pentagon Counter-UAS task force is developing lasers and electronic warfare
“We have a counter UAS, counter unmanned systems task force that's looking at everything from lasers, directed energy, which is one of my critical priority areas, to how do you do electronic warfare on these things to take them down.”
Emil Michael: Every frontier AI company will develop cyber capabilities adversaries will copy
“Cyber capabilities that are almost here, are certainly going to come from every frontier model company. At some point, it's certainly going to be you know, try to be distilled by the adversaries are going to be, you know, the next wave of innovation from these…”
Michael: Anthropic was first to aggressively pitch US government AI services
“They were the first to aggressively try to provide service to the government after the Biden administration's executive order about AI because they were, and again, you see this in the marketplace too, OpenAI was more focused on the consumer, with ChatGPT and …”
Michael: Tech-Pentagon Friction Stems from Mistaken Belief in Defense Safety
“The image that they might have, and this is not unique to them, of the Department of War or the administration, is that we don't have the safeguards that we do, that we're not paying attention to sort of the risks. We are, if not more so than most Americans wo…”
Michael: Anthropic contract banned AI use for weapons engineering and kinetics
“The provisions that were in the contracts that ultimately serve the Department of War said you can't use it for planning kinetic actions, you can't use it to develop weapon systems, so all the science, engineering, aerodynamics, all...”
Michael: Anthropic's autonomous weapons concerns were driven mostly by marketing
“The idea that autonomous weapons were an issue was sort of, I think, more marketing than anything. Because we have our own policy before that showed up that talks about that.”
Michael: Anthropic demanded Pentagon exceed mass surveillance laws in AI contracts
“They wanted us to rewrite the law because they thought Congress was just behind. They weren't understanding that new tech allowed new capabilities, but again, it's not our mission.”
Michael: AI vendors restricting kinetic use pose legitimate military supply chain risks
“If Lockheed Martin builds a weapon for me, and they're using, ah, a technology to help them do some of these science-oriented things, physics, aerodynamics, and so on, and the vendor has expressed an unwillingness to want that to be part of the use case, well,…”
Claude Mythos is first AI to complete end-to-end cyber network takeover
“This is from the AI Security Institute. This week we conducted cyber valuations of Claude Mythos preview and found that it is the first model to complete an AISI cyber range end-to-end, which means it's a 32 step corporate network attack from initial reconnais…”
Emil Michael: Relying on a single AI provider on classified networks was a mistake
“The original sin was In the past administration, choosing one AI provider and having no options, because it is a gargantuan effort to get these software things onto classified networks.”
Michael: AI labs' apocalyptic rhetoric justifies strict Pentagon scrutiny of management
“So these companies are talking about their things in apocalyptic terms, which make it necessary for us to judge the management teams, judge their actions,
Look at the terms of service.
Understand how they fit in our supply chain.
This technology is like nothin…”
Kantrowitz: Anthropic's defense models are hosted on Amazon's government cloud
“Anthropics models were hosted
On Amazon's cloud.
Their government cloud.
And so they upload the weights, weights to the model, and then you use it through Amazon.”
Kantrowitz: Anthropic cannot shut off Claude models hosted in third-party clouds
“Claude, I mean, Anthropic wouldn't have the capability to turn that off if it's hosted somewhere else.
Maybe upgrade it, in which case you flip them out, but to turn it off, they don't really have that capability.”
Emil Michael: AI model upgrade cycles have compressed to three months
“How many, the upgrade cycles for these things are now compressing to now three-ish months. So every three months you have a new set of model weights, a new set of guardrails, a new set of bugs in the way the model behaves, the way it hallucinates or doesn't ha…”
Michael: Anthropic's First Amendment retaliation claim against Pentagon will fail
“I think the notion that a First Amendment claim is going to hold up is, is, would be shocking, because that means that the government has no choice to make.”
Kantrowitz: Google's Gemini is available to military personnel on genai.mil
“There's a website, genai.mil, that's available to the people in the military here, and interestingly, Google, ah, is in there. Gemini's in there.”
Michael: Google personnel now regret withdrawing from Pentagon Project Maven contract
“If you fast forward from 2018 where the Google Maven thing happened to 2026, and you talk to people from Google who are involved in that, I think they regret it.”
Michael: US defense contractors consolidated from 50 in 1980s to five
“So, so in the eighties, during the height of the cold war, we had about 50 defense contractors, five zero, and they consolidated down to five.”