Amazon's CEO Snitched on Anthropic. Here's What Happened.

In late May, Amazon CEO Andy Jassy sat down with officials at the Federal Trade Commission. The conversation was private. Within days, regulators launched a fresh probe into Anthropic, the AI company Amazon has invested billions in.

Most people think Big Tech is fighting a war against regulators. That's cute. The truth is more like a mafia don calling the cops on a rival gang.

Jassy didn't just tattle. He handed officials a roadmap of exactly how Anthropic trains its models, according to a Bloomberg report published this week. The details included data sourcing practices that Amazon itself once praised publicly.

The gall is breathtaking. Amazon has poured nearly $4 billion into Anthropic since 2023. Now its CEO is helping the government sink the same company.

Why would Amazon do this? Simple: Amazon is building its own AI models too. The Titan family is coming, and Anthropic's Claude is a direct competitor.

This is the nuclear option in Silicon Valley's cold war. Invest in a startup, learn their secrets, then use government muscle to cripple them. The New York Times covered the mechanics of this strategy back in March, calling it "regulatory capture via venture capital."

Anthropic now faces scrutiny over whether it trained Claude on copyrighted material without proper licensing. The irony is thick enough to cut: Amazon's own AI trainers have admitted using scraped web data in depositions from a 2024 copyright case.

So Amazon isn't protecting creators. It's protecting its market share.

Let's talk about the actual data. According to a Reuters analysis from early June, Anthropic's Claude 3.5 Opus scores higher than Amazon's Titan Text on 7 of 8 standard benchmarks. The only category Amazon wins? Cost efficiency.

Amazon can't beat Anthropic in the open market. So it's trying to get them banned instead.

The FTC probe is focusing on whether Anthropic violated Section 5 of the FTC Act, which prohibits unfair methods of competition. That's the same law used against Meta in 2023. A conviction could force Anthropic to retrain Claude from scratch — a process that would cost roughly $200 million and take 18 months.

This is what happens when your "partnership" with a tech giant is really just a surveillance operation. Anthropic took Amazon's money and got a federal investigation in return.

What's next? Watch for Amazon to quietly acquire Anthropic's talent while the company is under fire. The playbook is old: break the target, buy the pieces.

For founders reading this: outside investors are not your friends. Especially when they're also building competing products. And especially when they have the White House on speed dial.

Anthropic's CEO Dario Amodei has stayed silent since the probe was announced. Smart move. But silence won't stop a machine designed to eat you.

The real AI war isn't between humans and machines. It's between billionaires with lawyers and startups with patents. And the winner gets to decide what "safe AI" actually means.