ChinaTalk Emergency Pod: Iran + Anthropic

Date: March 2, 2026
Host: Jordan Schneider
Guests: Emmy Probasco (CSET), Henry Farrell, Mike Horowitz, Brian Clark

Episode Overview

This "emergency pod" tackles two entwined crises: the implications of the U.S.-led military campaign in Iran—potentially the largest American air operation in decades—and the escalating conflict between Anthropic, a leading U.S. AI firm, and the Department of Defense (DoD). The conversation ranges from operational questions about air power and strategic aims in Iran, to the evolving role of AI (especially large language models like Claude) in military operations, and the political/economic fallout from the Pentagon’s standoff with Anthropic.

Key Discussion Points & Insights

I. The U.S. Air Campaign in Iran: Technology, Strategy & Leadership Decapitation

Precision Mass & The ‘Lucas’ System

• Mike Horowitz explains the use of America’s first “precise mass system”—the Lucas—ironically reverse-engineered from Iran’s Shahed drones, themselves copies of older Western tech.
  • “Effectively using Iran’s own technology against them.” [00:29]
• The operation is blending legacy U.S. strike systems (e.g., Tomahawk missiles) with new tech like Lucas and AI platforms (e.g., Claude).
• The scale and speed of operations represent a new era—over 500 strikes on the first day by Israel alone.

Leadership Targeting & Historical Rarity

• Jordan Schneider highlights the historical rarity and challenges of decapitating regimes via remote strikes without immediate follow-up or proxies to fill the void.
  • “A relatively rare thing in human history to pull off an assassination from a few thousand miles away.” [01:45]
• Brian Clark: Air power can eliminate leadership, but unless there’s civil society to fill the gap, or outside ground presence, chaos often results.
  • “Any good autocrat is going to eliminate any potential competition...so who picks up the pieces?” [03:08]

Strategy, Endgames & Risks

• “We’ve never seen the U.S. attempt a military operation of this scale with goals that were...this incoherent.” (Mike Horowitz) [06:09]
• Motivation appears mixed: regime change vs. simple threat elimination, with the Trump administration breaking from the “Pottery Barn Rule” (“if you break it, you buy it”).
• There is skepticism about long-term strategy:
  • “I’m not sure the Trump administration has a strategy beyond 2028.” (Mike Horowitz) [07:54]
  • “You have power vacuums...instability is likely...which increases your risk of terrorism, of militia groups lashing out.” [08:19]

Regional & Global Instability

• Gulf countries may bear the brunt of instability; they have strong incentives not to see Iran collapse.
  • “For the Gulf states, this is not great.” (Brian Clark) [09:40]
• Timing critique: If the aim was regime change, why not act when protests were strongest, rather than after mass casualties?
• Potential for increased nuclear proliferation in the region as a counter to perceived U.S. assertiveness.
  • “Everybody’s gonna get nuclear weapons now? Are you kidding?” (Mike Horowitz) [12:27]

Iranian Resilience and Protracted Conflict

• Iran has prepared for decades for distributed, decentralized warfare; significant command/control and missile/drone capability remain.
  • “They’ve been preparing for this kind of thing for decades...the IRGC survives and is trained to operate in a distributed way.” (Brian Clark) [16:43]
• The outcome is likely to be drawn out, with continuing threats to the Strait of Hormuz and Gulf security.

II. Operational Lessons: Air Power, Stockpiles, and Lessons for China

U.S. & Allied Stockpile Burnout

• Early media concern: The U.S. is rapidly burning through expensive interceptors to defend not just its forces, but also Gulf allies.
  • “If this keeps going, the U.S. will have to pull some stockpiles out of the Pacific and send them over to the Middle East.” (Mike Horowitz) [19:28]

Learning (and Showing) at Scale

• The U.S. is revealing much about its air defense operations to potential adversaries (notably China), as well as gaining valuable combat and operational experience.
  • “I wonder what information we are now communicating to China about how our air defenses operate and where the soft spots might be.” (Mike Horowitz) [21:09]
  • “This has given a lot of sets and reps...to evaluate the right defensive system...All of this is fantastic data for China.” (Brian Clark & Emmy Probasco) [22:37–23:22]
• U.S. naval and ground-based air defenses have “worked”—sometimes surprisingly well—but questions remain about cost efficiency.

III. AI on the Battlefield: The Claude & Anthropic Debate

“Claude” in Combat

• News broke that CENTCOM is actively using “Claude” (Anthropic’s LLM) integrated into the Maven Smart System (MSS) in Operation Epic Fury.
  • “They’re the most experienced with Maven Smart System...General Kurilla...did incredible work.” (Emmy Probasco) [25:21]
• Use cases are mostly mundane—data movement, workflow support, writing daily reports—that free up human focus for higher-level work.
  • “It’s really good at writing the daily report to the commander. There’s some really boring things that happen every day in an operation and AI can be helpful supporting people.” (Emmy Probasco) [28:38]
• The main value is giving commanders an edge in managing immense data, not direct battlefield lethality.

Structural & Bureaucratic Concerns

• LLMs and AI enable bureaucracies (especially huge ones like DoD) to function more efficiently—summarizing, translating, sanitizing intelligence, sifting data for foreign disclosure, etc.
  • “AI is fundamentally, in its current form, a bureaucratic technology.” (Henry Farrell) [29:05]
• There’s deep skepticism about “Terminator” or fully autonomous killer robots; military culture is fundamentally about control and redundancy.
  • “Military officers fundamentally like control...to cede so much control is not in their DNA or their training.” (Emmy Probasco) [32:03]

The Anthropic vs. Pentagon Dispute: Culture Clash & Policy Lesson

• The fight is characterized by clashing mental models: Pentagon wants treating AI like a weapons vendor (Lockheed); Anthropic prefers “service” logic, maintaining some say over “use cases.”
  • “This is a dispute about policy, but more about personalities and politics masquerading as policy.” (Mike Horowitz) [46:47]
  • “If you put ‘autonomous weapon’ in a contract, please define ‘autonomous weapon’—I’ll wait. It’s so hard.” (Emmy Probasco) [51:31]
• Anthropic doesn’t object to all defense work, but wants brakes on certain use cases (e.g., offensive autonomy).
• The DoD’s threat to designate Anthropic as a “supply chain risk” (a la Huawei) alarms panelists for its political and political–economic overreach.
  • “It was a remarkably stupid thing for the DoD to do.” (Henry Farrell) [56:02]
  • “If there’s even a non-zero chance you could get slapped with a designation that puts you in the same class as Huawei...what does that do to your incentives?” (Mike Horowitz) [59:55]

Strategic Fallout for U.S. Tech Policy

• This signals to Silicon Valley (and international partners) that working with DoD brings real reputational and legal risk, reducing the sector’s willingness to engage on national security contracts.
• U.S. supply chain rhetoric risks undermining its own “tech + democracy” alliance—both at home and abroad.
  • “If DOD wins, you’ll see allies and third countries looking at US tech companies like we look at Chinese tech companies.” (Henry Farrell) [57:40]

IV. AI, Autonomy, and National Security Bureaucracy

Policy, Training, & Systems

• Reliability and control—rather than raw capability—remain paramount in U.S. military adoption of new tech.
• Culture, training, and layered bureaucracy are presented as bulwarks against automation bias and accidental escalation.
• Research shows U.S. military personnel are less susceptible to automation bias than civilians, although not immune.
  • “Operators aren’t going to want to use systems that don’t work. Training and operating procedures matter.” (Mike Horowitz) [41:06]

Congress and Legislative Oversight

• Despite the drama, most of these are “boring” but essential bureaucratic/operational questions. Congressional debate (when it happens) is likely to be messy, populist, and not especially technocratic.
  • “I think we’re going to see a big debate on AI in Congress. It’s going to be a weird debate...not the technocratic debate that either the Pentagon or some people in Silicon Valley might expect.” (Henry Farrell) [75:00]

V. Geopolitics: U.S.-China Competition & Global Tech War

China’s Strategic Gains

• The panel highlights that the U.S.’s internal struggles—military overextension, Pentagon–tech sector tensions—create strategic opportunity for China.
  • “The winner in the Anthropic v. Pentagon feud is China.” (Mike Horowitz) [69:36]
  • “This is all good for...the Chinese. Who benefits from all of this? The Chinese.” (Mike Horowitz) [75:34]
• China is actively learning from watching the U.S. campaign in Iran (air defense, drone warfare, operational tempo) and is continuing its own AI/battlefield experimentation at pace.
  • “China continues to work. It’s hard at work. It’s frustrating to see all of this happening and think we’re having an argument that got out of hand as opposed to making real progress.” (Emmy Probasco) [78:27]

Political Economy & the “Tech Crackdown” Parallel

• The threat to blacklist Anthropic is compared, in tone if not in force, to Xi Jinping’s crackdown on Jack Ma—a signal of political control over the tech sector, though with less arbitrary power in the U.S. context.
  • “The actual powers that Trump or Hegseth have are not the same as Xi...But it’s terrible atmospherics for the U.S..” (Jordan Schneider) [77:12]

Notable Quotes & Memorable Moments

• On bombing as decisive policy:

  • “Bombing people doesn’t quite always get you what you want.” (Jordan Schneider) [01:45]

• On U.S. risk tolerance:

  • “They are willing to do things no American administration has ever been willing to do in the past because they don’t feel responsibility for governance over the places that they bomb." (Mike Horowitz) [06:24]

• On automation and AI:

  • “AI is fundamentally, in its current form, a bureaucratic technology.” (Henry Farrell) [29:05]
  • “Military officers fundamentally like control...to cede so much control is not really in their DNA.” (Emmy Probasco) [32:03]

• On U.S.-China rivalry:

  • “Who’s laughing at us right now? The Chinese.” (Mike Horowitz) [75:32]

• On the disorder between DoD and Anthropic:

  • “This is a pissing competition...about not simply egos, but who should be in charge of the world.” (Henry Farrell) [68:47]

• On the impact for democracy:

  • “The place it comes from is a fundamental notion that we are in a battle between democracy and authoritarianism and we’ve got to do everything we can to make sure that democracy wins.” (Henry Farrell) [66:59]

Timestamps of Important Segments

• 00:00–03:42: Opening, U.S. strike on Iran, military tech context
• 05:06–11:13: Efficacy and risks of leadership targeting, air power
• 12:25–15:33: Nuclear proliferation risk, Iran’s resilience
• 17:40–19:53: Operational stockpiles, Indo-Pacific implications
• 20:38–24:38: Lessons for China, air defense learning
• 25:03–29:02: Claude/Large Language Models in CENTCOM & military ops
• 29:05–34:14: AI as bureaucratic tech, automation bias, fail-safes
• 46:47–51:53: Anthropic-DoD policy fight—service vs. weapons logic
• 56:02–63:30: Political and economic implications of blacklisting Anthropic
• 65:08–69:36: Commercial incentive structures, patriotism, defense engagement
• 72:50–75:00: Congress and AI oversight
• 75:29–79:11: China, technology rivalry, concluding thoughts

Closing Reflection

The Iran crisis and the AI/Anthropic dispute are more than news events—they expose deeper, structural issues in U.S. strategic culture: an ambiguous theory of victory hampered by inadequate planning for aftermath; a tech sector–government relationship struggling to adapt to new realities; and a mounting challenge from authoritarian rivals abroad. If there is a common thread, it's the imperative for both the U.S. military and its broader tech-industrial base to clarify priorities, incentives, and processes—or risk ceding ground in an era where the margin for error is wafer-thin.

“While we’re having this argument, China is hard at work...It’s frustrating to see all of this happening and think that we are in a competition here and we’re having what seems like an argument that got out of hand as opposed to making real progress.”
— Emmy Probasco [78:27]

[End of summary]