Claude is the AI that said no.

Most AI companies will tell you they’re building technology for the good of humanity. Then they sign contracts with defense departments, hand their models over to governments, and let you figure out the rest.

Anthropic — the company behind Claude — did something different. When the Pentagon came knocking and asked them to remove the guardrails that prevent Claude from being used for mass domestic surveillance and fully autonomous weapons systems, Anthropic said no.¹

Not reluctantly. Not after a long negotiation where they caved on the important stuff. They walked away from a $200 million contract.²

What they refused to allow

The Pentagon wanted a single clause added: that Claude could be used for “any lawful use.” Sounds reasonable until you understand what “any lawful use” actually covers — including deploying AI in autonomous weapons systems that can select and kill targets without a human making that call, and using it to surveil American citizens at mass scale.³

Anthropic CEO Dario Amodei put it plainly: he “cannot in good conscience” agree to those terms.⁴

The Trump administration responded by directing all federal agencies to stop using Anthropic’s technology. The Pentagon labeled Anthropic a “supply chain risk” — a designation Anthropic is now fighting in federal court.⁵

That kind of retaliation tells you everything. When a company stands up to the government and the government tries to destroy them for it, that company was probably protecting something worth protecting.

What Claude actually uses, energy-wise

You’ve probably seen headlines about AI destroying the planet. There’s a real conversation to be had there — but it’s aimed at the wrong target when it’s aimed at you.

For personal and small business use, Claude’s energy footprint is relatively modest. A single response uses roughly 0.84 to 4 watt-hours depending on complexity — comparable to running your refrigerator for a few minutes.⁶ If you’re using Claude to help draft emails, research a topic, or run a small business, your cumulative AI energy use in a day is roughly equivalent to running a dishwasher once.⁷

That’s not a crisis. That’s a tool doing its job.

So why are data centers actually exploding?

Here’s what gets left out of most AI energy conversations: the data centers driving the most explosive growth aren’t powering your Claude chats. They’re powering something else entirely.

Amazon, Google, and Microsoft collectively invested roughly $750 billion in data center expansion in 2025 and 2026.⁸ Google’s electricity consumption surged 37% in a single year to 43.6 million megawatt-hours — the largest single-year spike in its history.⁹ Amazon’s data centers consumed 78,000 gigawatt-hours in 2025 alone.¹⁰

These are the same companies with CIA contracts, NSA partnerships, and multibillion-dollar deals with foreign governments.¹¹ Amazon has been the cloud provider of choice for intelligence agencies since the CIA’s first contract in 2013.¹² Google maintains a multi-billion dollar deal with a foreign government under Project Nimbus.¹³ Microsoft has quietly expanded its defense and intelligence footprint for years.

The energy demands being blamed on “AI” are largely the cost of building out planetary-scale surveillance infrastructure — not the cost of you asking an AI to help you write a letter or understand your privacy rights.

Why this matters

Choosing Claude isn’t just a practical decision. It’s a values decision. It’s choosing the company that walked away from $200 million rather than let AI target people without human oversight. It’s choosing an AI built by people who will take a real hit — government blacklists, lost contracts, federal retaliation — to hold a line.

Not every AI company can say that. Most of them can’t say anything close.

Use Claude