Altman Proposes Taxing Robots While Selling Them to the Pentagon

The company that dropped its ethics lost to the one that kept them

People think I’m a cheerleader for AI. I would say that I’m not a cheerleader for anything. I just want to look at issues logically.

AI is not good or bad – it’s just a tool – used by people – who decide how to use it.

And sometimes – HOW AI is used can be a revelation about the people who created it and the people who use it.

For example – let’s look at what’s happening with OpenAI – the company behind ChatGPT. They burned through $8 billion last year.

Eight billion.

Deutsche Bank said no startup in history has ever burned money like this. They spend about $1.69 for every dollar they bring in.

So this week Sam Altman published a 13-page policy paper about how America needs to tax robots and create a public wealth fund so everyone benefits from AI.

Four-day workweek. New social contract. The whole thing reads like a campaign platform for a guy who isn’t running for anything.

Tax the robots, share the wealth, work less.

Sounds great.

Except five weeks ago his company signed a classified deal with the Pentagon to put their AI models inside the military’s most secure networks. And the only reason they got that deal is because another company said no.

Anthropic – the other big AI company – refused to do mass domestic surveillance. Refused autonomous weapons. They had red lines and they kept them. So Defense Secretary Pete Hegseth did something that had literally never been done to an American company before. He designated Anthropic a “supply-chain risk.”

That’s a classification usually reserved for foreign adversaries. Chinese telecom companies. Compromised vendors. Not an AI startup in San Francisco that said “we’d rather not build a killbot.”

Anyway, a federal judge blocked it.

Judge Rita Lin called it an “Orwellian notion” that an American company could be branded a saboteur for disagreeing with the government. She wrote that the Pentagon’s own records showed it labeled Anthropic a risk because of its ‘hostile manner through the press.’ Textbook First Amendment violation.

But by then OpenAI already had the contract. Experts estimate it’s somewhere between $500 million and $2 billion over five years.

When there was a public outcry against OpenAI for taking the contract, and their employees started quitting, Altman admitted it “looked opportunistic and sloppy.” Said they “shouldn’t have rushed.” So they revised some of the contract language, added some principles about surveillance, and kept the money.

So – the timeline is – his company dropped its own ban on military applications back in January 2024, and in February he grabs a classified Pentagon contract. In April he publishes a paper saying the government should tax robots and help people whose jobs get destroyed by AI.

Basically – he’s asking the government to cushion the blow from his product while selling a weaponized version of it to the same government.

And the deal only happened because the competitor who said no got punished for it. The company that kept its ethics got branded a national security threat. The company that dropped its ethics got the contract.

I’ve written about OpenAI’s finances before. Their web traffic share dropped from 87% to 65% in the last year. Google ate 16 points of that. A 2025 report from the venture firm SignalFire found engineers were eight times more likely to leave OpenAI for Anthropic than the other way around. They’re hemorrhaging cash, hemorrhaging talent, and the grand plan is more data centers and military contracts. They also just raised 122 billion dollars and are planning for an IPO later this year.

But here’s the part that should probably keep Sam Altman up at night.

This week, Anthropic – the company the Pentagon branded a national security threat for having ethics – passed OpenAI in revenue. Thirty billion dollars in annualized revenue, up from nine billion at the end of last year. OpenAI is at twenty-five billion.

By their own numbers, the company that said no to the Pentagon tripled their annualized revenue in 3 months and is now bringing in more money than the company that said yes.

And it’s doing it while planning to spend about a quarter of what OpenAI does on training, by the companies’ own projections.

OpenAI is projected to burn $125 billion a year on compute by 2030. Anthropic’s projection for the same period is around $30 billion. Same race. Four times less money.

So, Altman raised $122 billion, reversed his ethics policy, took a classified military contract from a competitor that got blacklisted for having principles, and published a paper about how the government should protect people from AI – and his company is still losing. To the company that kept its red lines.

Both companies are reportedly getting ready to go public – Anthropic as soon as this year. OpenAI is reportedly eyeing a valuation of up to a trillion dollars. Analysts at PitchBook looked at both and concluded OpenAI scores weakest on business quality fundamentals despite having the highest valuation.

The company that said no to the Pentagon is now worth more to actual paying customers than the company that said yes. The ethics were the business strategy. And it worked.

That might be the most surprising sentence in American tech right now.

Sources

Correction, September 27th, 2026: This essay first said Altman reversed OpenAI’s ban on military use in February to get the Pentagon contract – the company dropped that ban in January 2024. It also said OpenAI lost $14 billion last year – it burned through about $8 billion in 2025, and $14 billion was a forecast for 2026; that it spends $3.30 for every dollar it brings in – it’s about $1.69; and that the policy paper came out last week – it came out on April 6th, the same week this essay ran.

Update, September 27th, 2026: A few lines now say who says so – the revenue figures are the companies’ own and the contract’s value is an outside estimate – and the supply-chain-risk label is described more exactly. Since this ran, the San Francisco court set that label aside in August – but on September 25th, a federal appeals court in Washington upheld, 2 to 1, a separate Pentagon order under a broader law that keeps Anthropic’s Claude out of the Defense Department’s supply chain. A few more lines now say who says so – the finding on the Pentagon’s records was the judge’s, the numbers on engineers leaving come from one venture firm’s 2025 report, the training-cost gap is the companies’ own projection, and the IPO plans and trillion-dollar valuation are what’s been reported.

Everything I have written about AI is in one place at rachelandthecity.com/the-ai-files.

The company that dropped its ethics lost to the one that kept them

Sources

Share this: