The Day AI Ethics Died: Inside OpenAI's $200M Pentagon Deal That Triggered a 295% Uninstall Spike

The Pentagon gives Anthropic a deadline to comply. Amodei holds firm.

February 27:** The deadline passes. Trump posts on Truth Social, calling Anthropic "Leftwing nut jobs." He orders all federal agencies to stop using Anthropic products immediately, with full phase-out within six months. Defense Secretary Pete Hegseth places Anthropic on the "national supply chain security risk" blacklist. [NP

Within hours:** OpenAI CEO Sam Altman announces a contract with the Department of Defense to supply ChatGPT for classified military networks. The deal is reportedly worth up to $200 million. [NPR]

The optics are devastating: one company stood up to political pressure and got destroyed; the other saw an opening and sprinted through it.

The Comparison: Two Companies, One Choice

The irony is almost too perfect: the Pentagon chose the technically inferior product because the better one had a conscience.

The Data Backlash: Users Voted with Their Thumbs

This wasn't just Twitter outrage. The numbers are brutal.

Uninstalls:** ChatGPT's mobile app saw a **295% spike** in uninstall rate on February 28. [BT citing Sensor Towe

Downloads:** New downloads dropped **13%** on February 28. [B

Ratings:** One-star reviews surged **775%**. Five-star reviews fell **50%**. [B

The #QuitGPT movement:** Organizers claim **1.5 million users** have pledged to delete ChatGPT or cancel subscriptions (movement's own figure, not independently verified). [B

Meanwhile, Claude's U.S. usage surged **51%** on February 28, and reached #1 in the U.S. App Store Productivity category on March 2. Claude's US downloads increased **240% month-over-month** in February. Claude's US downloads increased 240% month-over-month in February — pushing its usage far above normal levels. [BI]

This is what a user revolt looks like when it's backed by real data, not just hashtags.

The Internal Revolt: Even OpenAI's Own People Quit

On the same day the deal was announced, **Caitlin Kalinowski**, who joined OpenAI from Meta in 2024 to oversee hardware in its robotics division, resigned. She posted on X: *"AI has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got."* [BI]

When your own hardware lead-someone building the physical systems AI will run on-quits over a military deal, you have a problem that no PR spin can fix.

The Ultimate Humiliation: Pentagon Admits ChatGPT Is Worse

Here's the part that should keep Sam Altman up at night.

According to defense sources cited by [IC], Pentagon officials discovered within days that ChatGPT lacks the specialized capabilities needed for military applications. The system was simply not designed for the high-stakes, low-latency, multi-domain operations that Claude had been optimized for over years of defense collaboration.

In a twist that borders on farce, Pentagon officials had to **go back to Anthropic** and ask for help-because they couldn't easily remove Claude from existing weapons systems. [IC]

So the U.S. military blacklisted the company with the better product, then had to beg for its help after signing with the inferior alternative. This is not a strategy. This is chaos dressed up as patriotism.

The IPO Shadow: $500 Billion vs. $380 Billion-But at What Cost?

Both companies are planning IPOs in 2026. OpenAI was valued at **$500 billion** in its 2025 share sale. Anthropic stood at **$380 billion** as of February 2026. [IC], [BI]

Investors now face a stark choice:

The market will decide which bet pays off. But here's the question no one is asking: **If OpenAI is willing to sell out its safety principles for $200 million, what else is it willing to sell?**

Reverse-Thinking Judgment: The Industry's Foundational Lie

This isn't a story about two companies making different choices. It's a story about the **central contradiction of the AI industry**.

For years, every major AI lab claimed to be "mission-driven." OpenAI's mission was "safe AGI that benefits all of humanity." Anthropic's was "AI that is helpful, honest, and harmless." Both raised billions on that promise.

Now we know: **one of them meant it.**

OpenAI's behavior reveals that "mission-driven" was always a branding exercise, not a governance model. When the government said "remove the guardrails," Anthropic said no. When the Pentagon called with a $200 million check, OpenAI said "when do we start?"

The lesson is brutal: **Ethics in AI is not a feature. It's a liability.** The companies that take it seriously will lose contracts, get blacklisted, and face political retaliation. The companies that treat it as marketing will win deals, grow faster, and-if the user backlash fades-dominate the market.

This is not a criticism of Anthropic. It's a warning about the industry's incentive structure. As long as the U.S. government rewards compliance over conscience, the race to the bottom will continue.

Recommendations

For individual users:** If you value ethical consistency, switch to Claude. The usage spike is real, and the product is better for most knowledge work anyway. But understand that Anthropic's $380 billion valuation is itself a red flag-no company that large is pur

For developers:** Do not build on a platform that pivots on ethics. OpenAI's API could change its terms tomorrow based on political pressure. Anthropic has proven it will hold the line, even at great cos

For investors:** The short-term winner is OpenAI-it has the Pentagon contract and the higher valuation. But the user backlash is real, and technical inferiority in defense applications is a long-term liability. Anthropic's moral capital is an asset that cannot be easily replicate

For everyone:** Stop believing mission statements. Watch what companies do when the government asks them to break their promises. That's the only data that matter