Public transcript

Why the Claude AI Founders Told the Pentagon No

By: Oprah
Length: 9 min
Language: English
Text from: the video's captions
Transcribed: 14 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Anthropic's founders explain why they refused the Pentagon's request to remove Claude's safety guardrails for military use. They objected specifically to fully autonomous weapons and domestic mass surveillance, saying these conflicted with democratic values even as they supported national defense. Despite fearing it could end the company, the co-founders were unanimous, framing it as a policy rather than political decision, and they connect it to earlier principled choices like rejecting ads and publicly opposing a bill that would have banned all AI regulation.

Key points

Anthropic refused to remove safety guardrails for two military use cases: fully autonomous weapons and domestic mass surveillance.
The co-founders were unanimous in the decision, even though they believed it could be the end of the company.
The Pentagon had at times claimed it could stop other companies from doing business with Anthropic entirely.
Daniela described the conflict as a policy disagreement, not a political one, and noted Anthropic was the first AI company to work closely with national security.
Anthropic also rejected advertising because people share private health, financial and personal information with AI models.
In mid-2025, Dario wrote a New York Times op-ed against a bill that would have banned all AI regulation; it was voted down in the Senate 99 to 1.

Questions it answers

00:00The refusal

Why did Anthropic refuse to remove Claude's safety guardrails for the military?

They had worked with the Department of Defense but refused two use cases they believed went against the country's values: fully autonomous weapons and domestic mass surveillance.

Fully autonomous weapons like a drone army were off limits.
Domestic mass surveillance using government power to spy on Americans was off limits.
All other companies agreed to the Pentagon's terms.

02:02Unanimity and fear

Did the co-founders hesitate or fear for the company's survival?

Yes — it crossed their minds it could be the end of the company, but they felt it was the right thing to do, and the co-founders believed it unanimously.

The decision was unanimous among co-founders.
They reasoned that giving up their values would mean they weren't the company they started.

03:05Threats of retaliation

What retaliation did the Pentagon threaten?

At times the Pentagon claimed it could stop other companies from doing business with Anthropic entirely, including anyone connected to government business.

They could allegedly cut off all outside business relationships.
Everyone knew it was wrong, and they hoped others would side with Anthropic.

04:27Policy, not politics

How did Daniela frame the disagreement politically?

She saw it as a policy disagreement, not a political one, noting Anthropic was the first AI company to work closely with national security while holding these red lines.

Anthropic supports thoughtful government regulation of AI.
The company will work with governments, states, and NGOs.

06:01The no-ads decision

Why did Anthropic reject advertising revenue?

Because people have very private conversations with AI models — health, finances, children, personal challenges — and they wanted to treat customer data with the utmost privacy and respect, resisting pressure from investors who favored ads.

Ads create wrong incentives for AI.
Co-founders never questioned it, though investors pushed back.

07:35Opposing the regulation ban

How did Anthropic respond to the 2025 bill banning AI regulation?

Dario wrote a New York Times op-ed against a bill that would have banned all AI regulation; the bill was voted down in the Senate 99 to 1, and Anthropic was clearly on the right side.

Speaking out did cost some political relationships.
The bill lost 99 to 1 in the Senate.

Notes

The Pentagon Refusal

Anthropic had already worked with the Department of Defense and believes in defending the country, but not at the cost of the country's values.
Two use cases crossed the line: fully autonomous weapons (an AI-powered drone army with one finger on the button) and domestic mass surveillance (government spying on Americans).
The Pentagon did not agree, so Anthropic could not agree either, while other companies did.

Standing Firm as Co-founders

The co-founders met around a table and were unanimous; they thought it could be really bad, even the end of the company.
They asked whether their past selves would be proud if they compromised, and that made the decision clear.
The Pentagon had at times made claims it could stop other companies from doing business with Anthropic entirely, including anyone doing business with the government.
Daniela said Anthropic is about policy, not politics, and was the first AI company to work closely with national security, except in the two areas that could compromise democratic values.

The No-Ads Decision

People have very private conversations with AI models, sharing health information, finances, questions about their children, and personal challenges.
Co-founders refused ads despite quiet pressure from people inside the company and investors who pointed to Google and Facebook as the standard model.

Speaking Out on Regulation

In mid-2025, a bill in Congress would have banned all state regulation of AI and added no federal regulation either.
Advisers warned that speaking out would break political relationships and anger other companies, and some of that did happen.
Dario wrote an op-ed in The New York Times against the bill, and it was voted down in the Senate 99 to 1.

How Values Ladder Up

Oprah noted that big principled decisions are built from smaller ones along the way.
Smaller decisions like no ads and speaking out on regulation laddered up to the Pentagon moment.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.