Pentagon Blacklists Anthropic While Using Claude for Iranian Airstrikes

They banned the company for refusing mass surveillance. Then they kept using its AI to select 1,000+ daily strike targets.

TL;DR:

  • The Pentagon officially designated Anthropic a "supply chain risk" on March 5. First time a US company has received this label, usually reserved for adversaries like Huawei
  • The reason: Anthropic refused to drop guardrails preventing Claude from mass surveillance on Americans or autonomous weapons
  • The twist: The military is still using Claude in Iran. On Day 1 of operations, Claude generated approximately 1,000 strike targets with GPS coordinates, weapons recommendations, and automated legal justifications
  • Anthropic CEO Dario Amodei: "We do not believe this action is legally sound and we see no choice but to challenge it in court"
  • Consumer response: Claude downloads skyrocketed. Over 1 million new signups daily, making it the #1 AI app in 20+ countries

The Official Designation

On March 5, 2026, the Pentagon formally notified Anthropic that the company and all its products have been designated a supply chain risk, effective immediately.[1]

This is the first time a US company has received this label. It was designed for foreign adversaries: think Huawei, ZTE, companies whose technology might have backdoors for the Chinese government. Now it's being used against an American AI startup that refused to enable mass surveillance.

Defense Secretary Pete Hegseth announced the threat last week. The formal notification came this Thursday.

The designation means defense contractors and military suppliers must certify they don't use Claude in any Pentagon work. Lockheed Martin has already announced it will seek alternative AI vendors.[2]

What Anthropic Refused to Do

The dispute came down to two things Anthropic wouldn't give up:

  1. Mass domestic surveillance: Using Claude to analyze data collected on American citizens at scale
  2. Fully autonomous weapons: AI systems that identify and kill targets without human approval

Pentagon CTO Emil Michael argued these uses are already prohibited by law: "At some level, you have to trust your military to do the right thing."[3]

Anthropic CEO Dario Amodei wasn't buying it: "Safeguards relate to high-level usage areas, and not operational decision-making."

Translation: Law says one thing. Practice says another. Anthropic wanted it in writing.

The Part They Didn't Mention

Here's where it gets interesting.

At the same time the Pentagon was branding Anthropic a security risk for refusing to remove ethical guardrails, the military was actively using Claude to select strike targets in Iran.[4]

According to military sources, on the first day of operations alone, Claude generated approximately 1,000 prioritized targets. The system synthesized:

  • Satellite imagery
  • Signals intelligence
  • Surveillance feeds

Output: target lists with precise GPS coordinates, weapons recommendations, and automated legal justifications for strikes.

The system is embedded in Palantir's Maven Smart System on classified military networks, serving over 25,000 users across US Combatant Commands.

A military source told the Washington Post: "We're not going to let [Amodei's] decision-making cost a single American life."[5]

So the Pentagon's position is: we'll blacklist your company while continuing to use your AI to select targets for airstrikes. Just not for mass surveillance of Americans. That's the line.

Anthropic's Response

Amodei called the action "retaliatory and punitive."[6]

"We do not believe this action is legally sound, and we see no choice but to challenge it in court."

Anthropic's statement emphasized that applying the supply chain risk tool to "penalize a U.S. firm for declining to remove safeguards against mass domestic surveillance and fully autonomous weapons is a category error with consequences that extend far beyond this dispute."

The company has six months to phase out of military use. It plans to fight the designation in federal court.

Congressional Pushback

Senator Kirsten Gillibrand (D-NY), who sits on both the Senate Armed Services Committee and Senate Intelligence Committee, didn't hold back:[1]

"A dangerous misuse of a tool meant to address adversary-controlled technology. This reckless action is shortsighted, self-destructive."

The supply chain risk designation exists to protect America from foreign adversaries embedding backdoors in critical technology. Using it against an American company that refused to enable surveillance of Americans is, as Gillibrand put it, "shortsighted."

The Public's Response

People noticed. And they picked a side.

In the week following the supply chain risk threat, more than one million people signed up for Claude each day. The AI assistant surpassed OpenAI's ChatGPT and Google's Gemini to become the top AI app in more than 20 countries in Apple's app store.[1]

The Pentagon's attempt to punish Anthropic for ethical guardrails turned into the company's biggest marketing event.

What This Says About OpenAI

While Anthropic gets blacklisted, OpenAI secured a military agreement.[3]

OpenAI CEO Sam Altman previously said he shares Anthropic's "red lines" on mass surveillance and autonomous weapons. But OpenAI signed. Anthropic didn't.

The EFF has already called out the "weasel words" in OpenAI's Pentagon contract: the ban on "intentional" domestic surveillance leaves loopholes for mass data collection that "happens to" sweep up Americans.[7]

Anthropic wanted explicit protections. They got blacklisted. OpenAI accepted implicit ones. They got the contract.

What This Means

The Pentagon has established a new precedent: American AI companies that refuse to enable mass surveillance can be branded security risks alongside Huawei and ZTE.

At the same time, AI systems are being used to select strike targets in real-time warfare, with "automated legal justifications" built into the output. An algorithm is now generating reasons why bombing a target is legal.

This is the surveillance state and the military-industrial complex converging. The companies that draw ethical lines get punished. The ones that don't get contracts.

Anthropic is fighting back in court. The outcome will determine whether AI companies can maintain ethical guardrails or whether government pressure will force them all to comply.

References

  1. NPR - Pentagon labels AI company Anthropic a supply chain risk 'effective immediately'
  2. Bloomberg - Pentagon Notifies Anthropic It's Deemed Firm a Supply-Chain Risk
  3. CBS News - Pentagon formally designates Anthropic a supply chain risk amid feud over AI guardrails
  4. CNBC - Anthropic officially told by DOD that it's a supply chain risk even as Claude used in Iran
  5. WSWS - Claude AI has selected over 1,000 targets in the US-Israeli war against Iran
  6. TechCrunch - It's official: The Pentagon has labeled Anthropic a supply-chain risk
  7. EFF - Weasel Words: OpenAI's Pentagon Deal Won't Stop AI-Powered Surveillance

Published: March 7, 2026