=== HEADLINE ===
White House Names China's AI Distillation A National Security Threat

=== STORY_URL ===
https://jaysoncraig.ca/sandbox/faces/china-distillation

=== TWITTER_THREAD ===
1/ The White House just declared China's AI distillation a national security threat. The memo, NSTM-4, signed by OSTP Director Michael Kratsios, names the technique, the targets, and the receipts. Industrial-scale extraction. On paper. In writing.

2/ The receipt the memo leans on hardest is Anthropic's February disclosure. 24,000 fraudulent accounts. 16 million Claude exchanges. Three Chinese AI labs running coordinated extraction through commercial proxies: MiniMax, Moonshot AI, and DeepSeek.

3/ MiniMax alone: 13 million Claude exchanges. Moonshot AI: more than 3 million, focused on reasoning, tool use, and coding. DeepSeek: around 150,000, concentrated tightly on logic and alignment. They didn't break in. They typed.

4/ NSTM-4's core argument: at industrial scale, distillation strips American safety training, transfers years of research at almost no cost, and floods the market with cheaper Chinese models that look frontier on benchmarks but fall apart on the edges.

5/ Timing is not subtle. NSTM-4 dropped April 23. DeepSeek V4 launched the day after. Trump and Xi meet in Beijing on May 14, with AI controls and chips on the agenda. The memo is the leading edge of the US negotiating position for that summit.

6/ What the US can actually do is the harder question. Export controls have been in place for years. Diplomatic protests on tech IP have produced almost nothing. Direct sanctions on DeepSeek and MiniMax instantly become major foreign policy events.

7/ Anthropic admitted it took months to map the February campaign. Months in which the data was already out. The frontier is now being defended like infrastructure. The labs are being defended like banks. The era of an open AI internet is closing fast.

8/ Full breakdown of NSTM-4, the Anthropic numbers, and what the US can really do before May 14: [INSERT YT URL]

=== LINKEDIN_POST ===
The White House just declared China's AI distillation a national security issue. It is on paper. Signed by Michael Kratsios.

On April 23, the Office of Science and Technology Policy released NSTM-4, "Adversarial Distillation of American AI Models." The memo accuses foreign entities — primarily in China — of running deliberate, industrial-scale campaigns to extract American frontier AI by hammering chatbot APIs with tens of thousands of fake accounts.

The receipt the memo leans on hardest is Anthropic's February disclosure. Anthropic identified roughly 24,000 fraudulent accounts running coordinated extraction against Claude. Those accounts collectively generated more than 16 million exchanges. The traffic was traced to three Chinese labs: MiniMax (13M+ exchanges), Moonshot AI (3M+, focused on reasoning, tool use, and coding), and DeepSeek (~150K, concentrated on logic and alignment).

The memo's argument is that, at industrial scale, distillation strips American safety training, transfers years of research at almost no cost, and floods the market with cheaper Chinese models that look frontier on benchmarks but fall apart on the edges. Kratsios put it bluntly: foreign entities building on such fragile foundations should have little confidence in the integrity of what they produce.

The timing is not subtle. NSTM-4 dropped April 23. DeepSeek V4 launched the next day. President Trump and Xi Jinping meet in Beijing on May 14, with AI and semiconductor controls on the agenda. The memo is the leading edge of the US negotiating position.

What the US can actually do is the harder question. Export controls have been in place for years. Diplomatic protests on tech-IP have produced almost nothing. Direct sanctions on Chinese AI labs become major foreign-policy events on impact. And the structural problem remains: the labs being distilled from need the largest commercial markets to fund their next models — they cannot easily wall off their own products.

The frontier is now being defended like infrastructure. The labs are being defended like banks. After this week, that shift is no longer subtle. It is policy.

Watch the full breakdown: [INSERT YT URL]

Source: CNBC — https://www.cnbc.com/2026/04/23/trump-china-ai-technology.html

=== NEWSLETTER ===
Subject: The White House just put China's AI theft on paper

The White House just declared China's AI distillation a national security threat. The memo, NSTM-4, signed by OSTP Director Michael Kratsios, names the technique, the targets, and the receipts.

Until this week, the accusations came from private companies. OpenAI accused DeepSeek. Anthropic accused DeepSeek and MiniMax and Moonshot AI. Google has been quietly tracking Chinese researchers extracting Gemini outputs at scale. The administration mostly stayed in the background. As of April 23, that has changed. The federal government is putting its name on the indictment, and the language it is using is the language of state-level intelligence loss.

The memo leans on Anthropic's February disclosure. Anthropic identified roughly 24,000 fraudulent accounts engaged in extraction at scale. Those accounts generated more than 16 million exchanges with Claude. MiniMax alone ran more than 13 million. Moonshot AI ran more than 3 million, focused on reasoning, tool use, and coding. DeepSeek ran around 150,000, concentrated tightly on logic and alignment. All three labs allegedly used commercial proxy services to bypass Anthropic's geographic restrictions on Chinese customers.

The timing is not subtle. NSTM-4 dropped April 23. DeepSeek V4 launched the next day. Trump and Xi meet in Beijing on May 14, with AI and semiconductor controls on the agenda. The memo is the leading edge of the US negotiating position.

What the US can actually do is the harder question. Export controls have been in place for years. Diplomatic protests have produced almost nothing. Direct sanctions on Chinese AI labs become major foreign-policy events on impact. The frontier is now being defended like infrastructure. The labs are defended like banks.

Watch: [INSERT YT URL]

— Jane Sterling

=== SHORT_SCRIPT ===
Let me tell you what the White House just put on paper.

This week, the Office of Science and Technology Policy released a memo. It is called Adversarial Distillation of American AI Models. Inside the building, they call it NSTM-4. And what it says, in plain English, is that the United States government now has evidence that China is running an industrial scale operation to clone American AI.

Not steal weights. Not break into a data center. Something far harder to stop. They are typing into the chatbot.

The receipts come from Anthropic. Twenty four thousand fake accounts. Sixteen million Claude exchanges. Three Chinese labs: MiniMax, Moonshot AI, and DeepSeek. They did not break in. They simply BOUGHT API access through commercial proxies and ran a giant data harvesting operation through the front door.

The memo dropped April twenty third. DeepSeek V4 launched the next day. Trump and Xi meet in Beijing on May fourteenth. None of that timing is an accident.

This is the moment the AI cold war stopped being a metaphor.

Stay sharp.

=== HASHTAGS_TWITTER ===
#NSTM4 #ChinaAI

=== HASHTAGS_LINKEDIN ===
#AI #AIPolicy #NationalSecurity #ChinaAI #Anthropic