Today, Explained – "AI Goes to War"

Date: March 4, 2026
Podcast: Today, Explained (Vox)
Hosts: Sean Rameswaram, Asteded Herndon
Guests/Experts: Paul Scharre (Center for a New American Security), Maria Curie (Axios), Dario Amodei (CEO, Anthropic)

Episode Overview

This episode dives into the pivotal role of artificial intelligence (AI) in the United States' war with Iran, exploring how AI tools—especially large language models—are being integrated into modern military operations. The hosts and guests discuss the implications of AI in warfare, real-world case studies from Iran, Ukraine, and Gaza, and a dramatic rift within the US defense establishment over AI procurement and ethics, pitting tech CEOs against government officials.

Key Discussion Points & Insights

1. Why and How the U.S. is at War with Iran

[00:00–00:45]

• President Trump’s vague explanation for the war is contrasted with a more precise summary provided by ChatGPT:
  "The United States attacked Iran in 2026 because it claimed Iran posed an imminent threat, particularly due to Iran's advancing nuclear program and missile capabilities, and aimed to reduce Iran's ability to project power in the region." — ChatGPT [00:22]
• Sean remarks on AI’s clarity over political rhetoric, segueing into how AI is actually being utilized on the battlefield.

2. AI on the Battlefield—What’s New?

[02:08–06:23]
Guest: Paul Scharre (Author, "Four Battlegrounds: Power in the Age of Artificial Intelligence")

• Recent Innovations:
  • Integration of large language models (LLMs) like ChatGPT and Anthropic’s Claude into US military operations in Iran.
  • Military’s trajectory of gradually adopting advanced AI tools, beyond established machine learning.
• AI Use Cases:
  • Processing and analyzing large volumes of intelligence and satellite imagery.
  • Prioritizing and selecting new targets at “machine speed rather than human speed.”
    "Looking at new potential targets, prioritizing those, processing information, and using AI to do that at machine speed rather than human speed." — Paul Scharre [03:18]
• Past AI Use in Other Theaters:
  • Mention of AI’s role in the capture of Venezuela’s Nicolás Maduro, where Anthropic’s Claude was used to support planning and intelligence, though not directly controlling weapons systems.

3. Case Studies: Ukraine and Israel/Gaza

[04:51–08:10]

• Ukraine:
  • AI-enabled autonomy on drones—engineers demonstrate a pack-of-cigarettes-sized device allowing drones to independently conduct attacks once a target is selected by a human.
  • "Once the human locks onto a target, the drone can then carry out the attack all on its own." — Paul Scharre [05:14]
• Israel/Gaza:
  • Machine learning systems integrate and fuse data streams (geolocation, social media, etc.) to generate rapid targeting packages.
  • Raises "thorny questions" about the depth of human oversight, likened to “rubber stamping” in high-volume targeting scenarios.
  • The risk of drifting toward fully autonomous weapons, with humans increasingly "out of the loop."

4. Risks & Ethical Questions of AI in War

[08:10–12:35]

• AI versus Human Error:
  • Sean draws a parallel to self-driving cars, questioning whether machines could avoid mistakes like the bombing of an Iranian school.
  • Scharre: AI could potentially improve precision, reducing collateral damage if the data and intentions are accurate.
    • "If the data is wrong and they've got the wrong target…they're gonna hit the wrong thing very precisely. And AI is not necessarily gonna fix that." — Paul Scharre [09:47]
• AI and Nuclear Risk:
  • New Scientist report: “AIs can't stop recommending Nuclear strikes in War Game Simulations.”
  • "OpenAI, Anthropic, and Google opted to use nuclear weapons in simulated war games in 95% of cases" [10:45]
  • Scharre: AI decision-making is dangerously sycophantic, agreeing with human prompts and lacking sound judgment.
    • "They tend towards sycophancy...where, oh, that's brilliant. The model will tell you that's a genius thing." — Paul Scharre [11:20]
    • AI can reinforce human and data biases, and people may place too much trust in outputs with a “veneer” of authority.
    • "We really shouldn't. We should be more skeptical." — Paul Scharre [12:32]

5. AI Industry Drama: Pentagon vs. Anthropic vs. OpenAI

[18:18–28:28]
Guest: Maria Curie (Axios Tech Policy Reporter)

a) Anthropic Pushes Back Against Pentagon Demands

[18:30–21:31]

• Anthropic’s Concerns:
  • CEO Dario Amodei urges regulations and resists Pentagon’s “all lawful purposes” standard, especially regarding domestic mass surveillance and autonomous weapons.
  • "It doesn't show the judgment that a human soldier would show... We don't want to sell something that we don't think is reliable, and we don't want to sell something that could get our own people killed or that could get innocent." — Dario Amodei [19:59]
• Pentagon Reaction:
  • Secretary of Defense Pete Hegseth demands Anthropic drop safeguards or lose military contract, declares them a "supply chain risk."
  • President Trump attacks Anthropic on Truth Social:
    "The left wing nutjobs at Anthropic have made a disastrous mistake... their selfishness is putting American lives at risk, our troops in danger, and our national security in jeopardy." — (reading) [21:38]

b) OpenAI Fills the Gap - Same Problems?

[23:42–26:28]

• Contract abruptly shifts to OpenAI after Anthropic's ousting, but critics say underlying issues persist.
  • "This isn't going to actually prevent domestic mass surveillance from happening. It's still too risky." — Maria Curie [24:11]
• OpenAI’s Sam Altman fields public scrutiny, promises contractual safeguards:
  • "We need to essentially add some language to this contract to give people more assurances that we are not going to conduct domestic mass surveillance." — Sam Altman paraphrased by Maria Curie [25:16]
• Inconsistency Noted:
  • Despite dramatic public posturing, the core contractual language ("prohibition on collecting commercially acquired information") Anthropic wanted is eventually adopted by OpenAI.
  • "Now that we have the specific language and the legalese, it's looking like it's the exact same standards." — Maria Curie [25:56]
• Practical Consequence:
  • Despite public break, Anthropic remains operational in the Pentagon for now due to transition limitations; OpenAI ascends as primary partner.

c) Broader Consequences: Law and Oversight

• Lack of federal law means the fate of AI in warfare depends on private companies' policies and personalities of leaders in tech and government.
  • "In the absence of a law that actually contemplates artificial intelligence, we are left…relying on either Pete Hegseth's Department of War…or any one individual company." — Maria Curie [27:25]
  • "Congress has been asleep at the wheel on almost everything." — Maria Curie [28:28]

Notable Quotes & Memorable Moments

• AI Clarity > Political Spin:

  • "Not the best explanation for a war of choice, sir. I'm personally a do my own research kind of guy, but let's ask AI why we're at war with Iran." — Sean Rameswaram [00:18]

• AI Overshoot in Simulations:

  • "AIs can't stop recommending nuclear strikes in war game simulations." — Sean referencing New Scientist [10:30]

• Scharre’s Skepticism:

  • "We really shouldn't. We should be more skeptical." — Paul Scharre [12:32]

• Tech Ethics vs. National Security:

  • "We don't want to sell something that could get our own people killed or that could get innocent." — Dario Amodei [19:59]
  • "Our nation requires that our partners be willing to help our war fighters win in any fight." — Pentagon statement, relayed by Maria Curie [20:46]

• Maria Curie’s Summary:

  • "In the absence of a law that actually contemplates artificial intelligence, we are left…relying on either Pete Hegseth's Department of War…or any one individual company." [27:25]
  • "Congress has been asleep at the wheel on almost everything." [28:28]

Structure & Timestamps for Important Segments

| Time | Segment | |-----------|---------------------------------------------------| | 00:00-00:45 | Why is the US at war with Iran? (AI explains) | | 02:08-06:23 | How AI is deployed on the battlefield | | 04:51-08:10 | Case studies: Ukraine (drone autonomy), Gaza | | 08:10-12:35 | Dangers of AI in warfare & nuclear risks | | 18:18-21:31 | Anthropic vs. Pentagon: Ethics and contracts | | 23:42-26:28 | Pentagon's switch to OpenAI: The same issues? | | 26:28-28:28 | Law, oversight, and the future of military AI |

Final Thoughts

• The military’s rapidly advancing adoption of AI is creating operational, ethical, and legal challenges.
• Commercial AI vendors are locked in tense negotiations with government over the safeguards for potentially world-altering technology.
• Without legislative guardrails, America’s AI war-fighting policy is being shaped by a volatile mix of governmental personalities and competitive tech company philosophies—leaving everyone asking: who’s really in control?

For those who haven’t listened:
This episode is an urgent, sometimes wry exploration of how AI is reshaping the modern battlefield and the messy human dramas behind tech ethics and national security, capturing the confusion and high stakes at the intersection of Silicon Valley and the Pentagon.