MagicTools
AI TutorialsAugust 7, 20268 views4 min read

When the President Calls You Selfish: Dario Amodei's Pentagon Standoff


title: "When the President Calls You Selfish: Dario Amodei's Pentagon Standoff" slug: dario-amodei-pentagon-standoff-en summary: "Anthropic CEO Dario Amodei refused the Pentagon's demands for unrestricted AI access, got called 'selfish' by the President, and sat for a CBS interview hours later. Here are the positions he refused to compromise on." category: ai-tutorials tags: [Anthropic, Dario Amodei, Claude, AI safety, Pentagon, autonomous weapons] status: published locale: en source: authored

When the President Calls You Selfish: Dario Amodei's Pentagon Standoff

February 28, 2026. The Defense Secretary designates Anthropic a "supply chain risk to national security." Hours later, CEO Dario Amodei sits for a CBS interview and refuses to back down. The President posts: "Their selfishness is putting American lives at risk."

This isn't a story about a tech CEO. It's about a man who stood in front of the most powerful military on Earth and said: not without limits.


What Happened

Defense Secretary Pete Hegseth delivered a three-day ultimatum: accept unrestricted military AI use, or get blacklisted. Anthropic refused. Hegseth followed through.

Hours later, Dario sat down with CBS's Jo Ling Kent. Calm. Direct.

The fight isn't technical. It's about where principle draws the line.


Two Non-Negotiables

Dario said Anthropic supports 98-99% of military use cases. Intelligence analysis. Cyber operations. Combat support. Already deployed across the intelligence community.

Two things remain off limits.

Red Line #1: Domestic Mass Surveillance

"Private companies collect data → government buys it → AI analyzes at scale. This isn't illegal. It just wasn't useful before AI." Dario explained it plainly.

AI moves faster than law. That gap is the problem.

Red Line #2: Fully Autonomous Weapons

He meant weapons that fire with zero human involvement. The reasoning splits two ways:

The tech doesn't work yet. "Anyone who's used AI models knows there's a basic unpredictability we haven't solved."

Nobody's designed oversight for it. A drone army with no human decision-makers — who decides what to target? We haven't had that conversation.


The Pentagon's Play

Trump called Anthropic's position selfish. Dario's response: "We wanted to strike a deal from the beginning."

Then he dissected the Pentagon's tactics. Language like "if the Defense Department deems it appropriate" and "in line with laws" created an appearance of compromise while leaving everything unchanged. The Pentagon spokesman tweeted the identical position publicly.

Classic institutional maneuver: agree to nothing while sounding reasonable.


Why This Isn't an Anti-Military Story

Anthropic is the most government-forward AI company in the space:

  • First to deploy on the classified cloud
  • First to build custom national-security models
  • Already running cyber ops and combat support

Dario: "I believe we must defend against autocratic adversaries like China and Russia."

He's willing to give the Pentagon the most powerful AI systems made. Just not without guardrails. Not for autonomous killing. Not for mass surveillance.


What The Stakes Actually Are

Dario kept returning to one phrase: "democratic values."

His framing: defend against autocrats, but don't become them in the process. If you destroy your democratic principles while defending them, you lose a different war.

He's not speaking in abstractions. Anthropic lost contracts. The President attacked him publicly. He got blacklisted.

A CEO who paid the price for principle — in real time, not in memoir.


✨ Original source: CBS News exclusive interview transcript, February 28, 2026.

Related Articles

The Agent Leaderboard Just Got a New King. Here's Why You Shouldn't Care.

Qwen3.8 Max tops the Agentic Index while Prime Agent open-sources a self-improving framework. Together they expose a crack in how we evaluate AI: you can't score an agent that rewrites itself. Your real workload is the only benchmark that counts.

developerAug 7, 20263 min
26

claude install Creates a Broken %h Symlink? Clean Debian Test Comes Out Fine (Not Reproduced)

A Fedora user reports on GitHub that the official one-line installer leaves ~/.local/bin/claude as a broken symlink pointing at the literal %h/.local/share/claude/versions/2.1.220 — the %h placeholder never expanded to the home directory. We ran the exact same install command in a clean Debian 12 container: the resulting symlink is correct (%h properly expanded), and install.sh now ships 2.1.223 (the report was against 2.1.220). This is a not-reproduced field report: full environment, commands, and raw output are included, along with three candidate explanations and a one-minute manual fix if you are currently stuck on the broken link.

claude-codeclaude-code-lab+4
claudeAug 7, 20262 min
24

Claude Code Says No conversation found to continue — Your -p Sessions Are Being Filtered Out

You run a headless claude -p command, then type claude --continue to pick it up interactively — and get No conversation found to continue, even though the session file sits right there on disk. This is a regression introduced in v2.1.90: the --resume picker's 'hide -p/SDK sessions' filter was mistakenly applied to --continue, which should continue the latest session unconditionally. We reproduced it end to end on macOS + v2.1.220 (the GitHub issue reports Fedora, so both platforms confirm), and verified two workarounds: headless -p --continue has always worked, and prefixing CLAUDE_CODE_ENTRYPOINT=sdk-cli lets interactive --continue bypass the filter with full history loaded.

claude-codeheadless+4
claudeAug 6, 20263 min
30

Which Model Is Claude Code Actually Using? settings.json vs ANTHROPIC_MODEL vs --model, Tested

Multi-agent setups, CI, and batch scripts all rest on one assumption: the session actually runs on the model you configured. But the model can be set in four places — settings.json, the ANTHROPIC_MODEL environment variable, the --model flag, and /model in-session — and the docs never give you one table saying which wins. Meanwhile a GitHub issue reports settings.json silently ignored on Windows, with a full day of work run on the wrong model and discarded. Using modelUsage as hard evidence, we tested the chain layer by layer: --model > ANTHROPIC_MODEL > project settings.json > built-in default, with the [1m] suffix form working too — plus a verification method more trustworthy than any UI hint.

claude-codeclaude-code-lab+4
claudeAug 6, 20263 min
30

Published by MagicTools