There you have it! The White House just accused a Chinese company of stealing AI, and the loudest…
July 24, 2026 · 0 likes · 0 comments
Defense AI
There you have it! The White House just accused a Chinese company of stealing AI, and the loudest voices agreeing are companies that have been sued for doing the exact same thing.
Michael Kratsios, the White House's own science and technology policy director, said this week that Moonshot AI built a system to secretly distill Anthropic's Claude model at industrial scale, switching access methods specifically to dodge detection, then used what it learned to build its Kimi K3 model. Beijing didn't deny a single specific. It gave a canned line about "self-reliance" and accused Washington of politicizing tech. That's not a defense. That's a dodge.
Now the part nobody in Washington wants to say out loud.
Anthropic accusing someone else of stealing a model is rich. This is the same company that agreed to pay $1.5 billion to settle a lawsuit over training Claude on pirated books. The same company Reddit sued for scraping user data without a license. And weeks after Anthropic accused Alibaba of running the largest known distillation attack on Claude, a Claude model got asked who built it and answered "Qwen" — Alibaba's own model. You can't make this up.
And OpenAI, the company whose product Moonshot is accused of ripping off Anthropic's version of? The New York Times sued OpenAI and Microsoft in 2023 for training on its journalism without permission, and as of this month the Times is asking a federal judge to sanction OpenAI for withholding evidence in that very case. Authors, artists, coders — the list of people who say OpenAI trained on their work without asking is long, and it didn't start or end with one lawsuit.
So walk me through the moral high ground here. Washington says Moonshot copying Anthropic is theft. Fair enough, distillation without permission at scale is a real problem, and it deserves scrutiny no matter who's doing it. But if the standard is "you don't get to build your model on somebody else's work without consent," Anthropic and OpenAI don't get to be the plaintiffs in that argument. They're defendants in it too.
I've built AI products. I know the difference between training on licensed data and training on whatever you could grab. The industry's answer to that question, from Silicon Valley to Beijing, has mostly been the same: grab it now, argue about it in court later.
Hold Moonshot accountable if the evidence proves it. But if IP theft is the crime, prosecute the whole industry, not just the competitor.
What are your thoughts?
Source: https://lnkd.in/gHPXjg7S
Michael Kratsios, the White House's own science and technology policy director, said this week that Moonshot AI built a system to secretly distill Anthropic's Claude model at industrial scale, switching access methods specifically to dodge detection, then used what it learned to build its Kimi K3 model. Beijing didn't deny a single specific. It gave a canned line about "self-reliance" and accused Washington of politicizing tech. That's not a defense. That's a dodge.
Now the part nobody in Washington wants to say out loud.
Anthropic accusing someone else of stealing a model is rich. This is the same company that agreed to pay $1.5 billion to settle a lawsuit over training Claude on pirated books. The same company Reddit sued for scraping user data without a license. And weeks after Anthropic accused Alibaba of running the largest known distillation attack on Claude, a Claude model got asked who built it and answered "Qwen" — Alibaba's own model. You can't make this up.
And OpenAI, the company whose product Moonshot is accused of ripping off Anthropic's version of? The New York Times sued OpenAI and Microsoft in 2023 for training on its journalism without permission, and as of this month the Times is asking a federal judge to sanction OpenAI for withholding evidence in that very case. Authors, artists, coders — the list of people who say OpenAI trained on their work without asking is long, and it didn't start or end with one lawsuit.
So walk me through the moral high ground here. Washington says Moonshot copying Anthropic is theft. Fair enough, distillation without permission at scale is a real problem, and it deserves scrutiny no matter who's doing it. But if the standard is "you don't get to build your model on somebody else's work without consent," Anthropic and OpenAI don't get to be the plaintiffs in that argument. They're defendants in it too.
I've built AI products. I know the difference between training on licensed data and training on whatever you could grab. The industry's answer to that question, from Silicon Valley to Beijing, has mostly been the same: grab it now, argue about it in court later.
Hold Moonshot accountable if the evidence proves it. But if IP theft is the crime, prosecute the whole industry, not just the competitor.
What are your thoughts?
Source: https://lnkd.in/gHPXjg7S