Inside the Iran War and the Pentagon's Feud with Anthropic with Under Secretary of War Emil Michael

Emergency pod vibes, episode 263 of All-In. We’ve got Emil Michael, now Undersecretary of War for research and engineering under Pete Hegseth, here for a rapid brief. Quick table-set: the U.S. and Israel have launched joint strikes on Iran under Operation Epic Fury; Iran’s supreme leader and dozens of senior officials are reported dead; a U.S. sub torpedoed an Iranian ship; six U.S. reservists were killed in Kuwait; and betting markets see real odds of boots on the ground and regime collapse this year. Emil, how long does this run, what’s the end state, and do we put people in country?

This is weeks, not months. The goal is to strip Iran’s ability to fund and arm proxies, degrade missiles and drones, and constrain the nuclear program, with no repeat of long, open-ended occupations.

Many saw Trump as the peace-time president; now we’re in multiple theaters. Freeberg, why this push now?

Think leverage. Big China meetings are coming, and Iran and Venezuela feed China’s energy needs. Hitting those flows before talks gives the U.S. bargaining power on trade, tech, and Taiwan without chasing regime change.

Zooming out, China’s growth is sliding and domestic pressure is rising, which can tempt a Taiwan adventure. Cutting off oil from Iran and Venezuela, and even narrowing Russia’s flows, hems China in and sets up a grand bargain that avoids a catastrophic war.

To calibrate the scale: we inherited Gaza and Ukraine; Houthis were a narrow freedom-of-navigation issue; Venezuela was a single, surgical night. The doctrine now is decisive action without long deployments.

These ops looked unusually clean. What changed under the hood?

Commanders who fought the post‑9/11 wars now run the show, and we’ve integrated space, air, sea, land, and cyber with mature planning and tech. War games were on the shelf, tactics kept current, and missions are built to minimize casualties and get out fast.

So is it tech gains or rule changes that unlocked this? And how do drones shift the fight ahead?

Both. We relaxed punitive rules of engagement and returned judgment to commanders with clear red lines. Drones have evolved from Predators to swarms and robot‑on‑robot fights, with AI doing target recognition and kill‑box search when links get jammed.

AI can err or hallucinate. How do you keep it safe enough for lethal contexts?

We’re not at a full self‑driving moment. Use AI where civilian risk is low, like autonomous lasers against drones or space intercepts against hypersonics, and tighten human oversight as density rises. Tactics are scenario‑specific and mature over the next decade.

Is there a live debate about handing kill decisions to machines?

Full autonomy for lethal decisions isn’t on the table. We’re at basic autonomy in drones, undersea vehicles, and collaborative wingmen. Also, if adversaries rip models and strip guardrails, we need models that can meet them lawfully and effectively.

Inside the right, there’s a charge that Israel drives U.S. moves. Is the president captured by that interest?

No. He acts on U.S. interests and leverages allied capability. Israeli intelligence is valuable, but the better question is why Europe hesitated while we and partners executed.

Status of Israel’s laser defenses, and can directed energy become a broad nuclear and missile shield?

Early versions work and are improving. The end state is layered defense—lasers and high‑power microwave at range, plus ground systems for cruise missiles and drones. It’s joint innovation; both countries have pieces.

On values, I’m pro freeing people from dictators when it can be done cleanly, and I trust this team to judge that window.

We’ve learned from Iraq and earlier Iran chapters; focused ops can avoid forever wars. If Venezuela and Iran flip back to freedom, that’s a historic advance for democracy.

Let’s hit the economic ripples. Oil, the Strait of Hormuz, and insurance markets are flashing red. Freeberg?

Hormuz risk spiked war premiums from a sliver of ship value to several times that, then many underwriters paused coverage. The U.S. stepped in with DFC to backstop war risk so tankers keep moving, which could seed a new U.S. maritime insurance market long dominated by Europe.

Breaking: Anthropic has been flagged as a supply chain risk after a canceled contract. Emil, what happened?

Their terms barred planning kinetic operations and other lawful military uses. We spent months pushing for an all‑lawful‑use standard. After Venezuela, they probed whether their stack was used, raising the specter of a mid‑fight shutdown. That dependency is unacceptable, so we moved to unwind and set a clear line.

Technically, could Anthropic interfere with your usage?

Yes. The model sits in GovCloud with Palantir brokering access, but Anthropic holds the control plane and can change behavior.

Lesson for everyone: go multi‑model. Philosophical whiplash can de‑platform your core workflows, and token‑hungry tools can blow up costs.

Costs will climb as use cases deepen, so redundancy and price pressure are key. Grok is in for all lawful use, Google is moving there on classified, and OpenAI engaged constructively. Google’s cloud stack is an economic edge.

Why formally label them a supply chain risk instead of just walking away?

We can’t have biased or controllable models designing weapons or components anywhere in our defense ecosystem. Use them for commercial planes if you want, but not for fighters and missiles.

They claim this is pay‑to‑play politics. Is that real, and why the public spat?

That’s nonsense. I tried for months to get to a simple, lawful‑use standard. They reframed it as mass surveillance and killer robots—neither reflects our legal limits or the real risk, which was reliability and control.

Fast fix: apologize, commit to all lawful use, and lock stable terms. That’s how you regain trust.

That’s exactly what we need—reliable partners who won’t wobble mid‑mission. But when you also publish broadsides, you owe more than a quiet tweak.

Last one for now: can we onshore critical defense supply chains fast enough?

We’re early but moving. With the Office of Strategic Capital, we’re lending against real needs—critical minerals, batteries, solid rocket motors, composites—to reduce China dependence. No immediate munitions shortfall, but we must plus up and modernize beyond Cold War kits.

Defense tech venture funding is approximately three times last year, and to keep that flywheel spinning we need startups winning big, fast contracts because war is shifting from a few giant carriers to lots of smart, low-cost systems that new entrants can build.

We keep hearing this will consolidate into new primes, with a handful like Anduril and Palantir capturing most of the value. Is that where this market lands?

They would love that, but I want a hungry second tier constantly pushing them; I just met a thirty-person shop shipping roughly a thousand E-RAM missiles a year for Ukraine, and I want those builders thriving while we strip away bureaucracy as a moat the old primes used to dominate contracting.

How do you unwind that bureaucracy so product innovation actually reaches you?

We are rewriting requirements to state the mission problem instead of bloated specs, inviting solutions with minimal must-haves, and shifting to commercial-style fixed-price deals with real incentives and penalties; startups lean into that, and I am driving it every day.

So you could say the Strait of Hormuz must stay open, describe the outcome, and let entrepreneurs pitch creative ways to achieve it?

Exactly—focus on outcomes and let builders compete.

You also oversee DARPA. What breakthroughs should listeners know about?

DARPA still attracts the best problem solvers; publicly, we are using biology to produce critical minerals without dirty refining to leap ahead of China, and we are hardening against AI-driven cyberattacks, with much more under wraps.

Two quick ones before we close. Aliens—and where does China truly stand relative to us?

I tried to get the alien brief and struck out; on China, their buildup over about fifteen years has been massive while we were focused elsewhere, but our submarines, space layer, and operational skill remain unmatched and we need to keep that lead from shrinking.

We cannot relax; it helps to know you are on it and that funding is flowing, but complacency is not an option.

This week was a reminder of how fortunate we are to have layered defense when civilians are under fire in places like Dubai, Doha, and Tel Aviv; even as a small-government person, I see defense as the foundation of our freedom, and I am grateful for what you do.

Thanks for being candid and thoughtful.

Such an insightful episode. We will see you next time.

Our playful, open-sourced theme kicks in and we riff as the music rolls.

We joke that we should all get a room for one big hug to release the tension.

More goofy ad-libs and chanty send-offs to close it out.