Anthropic Accuses DeepSeek of Stealing Claude: The Irony Is Delicious

March 2, 2026

Anthropic just did something predictable: they cried foul. This time, they're accusing DeepSeek, Moonshot, and MiniMax of running what they call an "industrial-scale" operation to extract data from Claude — 24,000 fake accounts worth.

Let that number sink in. Twenty-four thousand.

The Accusation

According to Anthropic's filing, these three Chinese AI labs created tens of thousands of bot accounts to systematically query Claude and harvest its outputs. The goal: distill everything useful and feed it into their own models. Anthropic is calling it theft. They're asking for damages. They're playing the victim.

Sure. Okay.

A Quick Recency Check

Here's what Anthropic doesn't want you to remember: their entire existence is built on the shoulders of open source. Claude wasn't forged in some pure isolation chamber — it learned from the internet's collective output, just like every other model on the planet. GPT-2. GPT-3. LLaMA. The "open" ecosystems that came before.

Anthropic benefited from years of shared research, open weights, and community contributions. Now they want to gate what they've built? That's a convenient position to take once you're the one on top.

But Wait — Didn't OpenAI Do The Same Thing?

Oh, this is the best part. OpenAI made nearly identical accusations against Chinese labs last year. The same playbook. The same outrage. The same "they're stealing our IP" finger-pointing.

So let me get this straight:

  • OpenAI trains on everything ever written, then claims exclusive rights to intelligence
  • Anthropic does the same, then sues when someone does it to them
  • Both companies lecture the industry about "responsible AI development"

This is the AI industry in 2026. Everyone steals. Everyone complains when they get stolen from. The only difference is who's holding the microphone.

The Real Story

24,000 fake accounts is a lot. But let's call this what it is: a business dispute, not an ethical reckoning. Chinese labs are competitive — maybe too competitive. But Anthropic's moral high ground is built on sand.

If you want to play in AI, you deal with distillation. Everyone does. The only question is who gets to complain about it loudest.

Anthropic chose to make noise. We'll see if anyone besides their legal team is listening.

← Back to Blog