A bombshell White House memo reveals that foreign entities, primarily in China, have been running industrial-scale, coordinated campaigns using tens of thousands of proxy accounts and jailbreaking techniques to systematically extract America’s most advanced AI breakthroughs. Washington is now fighting back.

The Bombshell Memo: What the White House Just Revealed

This is not a routine government policy update. This is one of the most significant national security disclosures in the modern AI era — and frankly, it’s something many of us in the national security and tech reporting space have been watching build for years.

On April 23, 2026, the Executive Office of the President — specifically the Office of Science and Technology Policy (OSTP) — issued a memorandum designated NSTM-4, titled “Adversarial Distillation of American AI Models.”

It was signed by Michael J. Kratsios, Assistant to the President for Science and Technology and Director of the OSTP, and addressed to the Heads of all Executive Departments and Agencies.

This memo is addressed to every corner of the federal government. That tells you everything you need to know about how seriously Washington is taking this threat.

What Is 'AI Distillation' — And Why Should You Care?

Before we go further, let me break down the technical term at the heart of this story, because it matters enormously.

AI distillation is the process of training a smaller, more cost-efficient AI model by feeding it the outputs of a larger, more advanced one. Done legitimately — as U.S. companies do — it is a tool for innovation: creating lighter, faster models without requiring the processing power of a supercomputer.

Done maliciously? It is digital theft at a scale we have never seen before.

Here’s how the adversarial version works: foreign operatives flood American AI systems — platforms like OpenAI’s ChatGPT, Anthropic’s Claude, and others — with millions of carefully crafted queries. By harvesting the responses at massive scale, they build enormous datasets that allow them to reverse-engineer and replicate the core capabilities of systems that took American researchers years and hundreds of billions of dollars to build.

And they do it at a fraction of the cost. That’s the theft. That’s the national security crisis.

Inside NSTM-4: What the Memo Actually Says

I’ve read every word of the NSTM-4 memorandum so you don’t have to. Here is what the U.S. government is officially putting on record:

The memo identifies the following specific threats:

• Foreign operatives are deploying tens of thousands of proxy accounts to evade detection while querying American AI systems at massive scale.
• They are using sophisticated jailbreaking techniques to bypass safety guardrails and extract proprietary, restricted outputs that would not otherwise be accessible.
• The resulting stolen models are being used to deliberately strip security protocols — including the ideological neutrality and truth-seeking mechanisms built into American AI systems.
• These stripped-down versions are then potentially fed into military, intelligence, and mass surveillance systems — enabling authoritarian governments to weaponize American innovation against their own people and against the United States.
• While the stolen models don’t fully replicate the originals, they are close enough to “perform comparably on select benchmarks at a fraction of the cost,” giving adversaries a massive, unfair competitive advantage.

This Isn’t Just Theory — Companies Have Already Been Hit

The reason NSTM-4 carries so much weight is that it isn’t speculative. American AI companies have already documented these attacks. In February 2026, Anthropic — the company behind the Claude AI — revealed that Chinese AI labs including DeepSeek, Moonshot AI, and MiniMax used approximately 24,000 fraudulent accounts to extract millions of Claude responses for use in training competing systems.

OpenAI has similarly reported large-scale distillation attempts. And in 2024, federal prosecutors charged a former Google engineer with stealing AI trade secrets and transferring them to Chinese companies — a criminal case that sent shockwaves through Silicon Valley.

Let that sink in. These aren’t isolated incidents. This is a coordinated, systematic, government-level campaign to dismantle America’s most valuable technological competitive advantage — one query at a time.

The Trump Administration’s Four-Point Counteroffensive

NSTM-4 is not just a warning — it is a declaration of action. The Trump Administration has formally committed to four concrete steps to protect American AI dominance:

1. INTELLIGENCE SHARING: Federal agencies will share active intelligence with U.S. AI companies about specific foreign attempts to conduct unauthorized distillation — including the exact tactics being used and the actors involved. No more leaving the private sector in the dark.
2. PRIVATE SECTOR COORDINATION: The administration will enable American AI companies to better coordinate their defenses — building a unified shield across the industry rather than leaving each company to fight alone.
3. BEST PRACTICES DEVELOPMENT: The White House will partner with private industry to develop the tools, protocols, and best practices necessary to identify, mitigate, and remediate industrial-scale distillation attacks before they succeed.
4. ACCOUNTABILITY MEASURES: The administration is exploring a range of measures to hold foreign actors accountable. Experts have floated export controls, diplomatic protests, and targeted technology restrictions. Make no mistake — consequences are coming.

The Geopolitical Flashpoint: The Trump-Xi Summit

Here is what makes this memo even more explosive in its timing: it was released just three weeks before President Trump is scheduled to meet with Chinese President Xi Jinping in Beijing on May 14, 2026 — a summit that was already delayed once from its original late-March date.

Releasing NSTM-4 now is not accidental. This is a deliberate geopolitical signal. Washington is walking into Beijing with documented evidence of AI theft and a public declaration that it will no longer tolerate it. This memo will almost certainly be on the table — openly or otherwise — when Trump and Xi sit down.

The Chinese Embassy in Washington has not immediately responded to requests for comment on the memo. That silence is telling.

Kratsios Draws a Hard Line — And Issues a Warning

Director Kratsios acknowledged that legitimate distillation plays a vital role in a healthy AI ecosystem. But he drew an unmistakable line:

Kratsios also issued a sobering strategic warning to foreign entities banking on their stolen AI foundations: as detection and mitigation technologies improve, those who built their AI capabilities on surreptitious distillation should have “little confidence in the integrity and reliability of the models they produce.” In other words — your stolen models have an expiration date.

Why This Matters for Every American

I know some of you may be thinking: “This is about tech companies and government — what does it have to do with me?” I want to answer that directly.

America’s AI lead represents decades of foundational scientific research, hundreds of billions of dollars in private investment, and the creative genius of hundreds of thousands of American engineers, mathematicians, and entrepreneurs. That lead drives the jobs of tomorrow. It drives breakthroughs in medicine, in climate science, in national security, in education. It is, in every sense, a national treasure.

What China is doing — if the White House evidence holds — is stealing that treasure systematically, at industrial scale, without an ounce of the innovation, investment, or risk that produced it. They want America’s future. And they want it for free.

That affects every American. Full stop.

NSTM-4 At-A-Glance: Key Facts From the Memo

NSTM-4 is a watershed moment. For the first time, the Executive Office of the President has formally put the U.S. government on record with evidence of industrial-scale, coordinated AI theft by foreign entities — and issued a direct commitment to fight back.

Watch for escalating tensions at the May 14 Trump-Xi summit in Beijing. Watch for potential export controls and economic countermeasures. Watch for legal actions against specific companies and individuals. Watch for new AI security requirements for American firms.

The AI war is real. It has been fought in the shadows for years. Now it’s in the open — and America is declaring that it intends to win.

About the Author

Samuel A. Lopez is an investigative correspondent for USA Herald. A nationally recognized journalist covering national security, constitutional law, federal politics, and the intersection of technology and American sovereignty, Lopez has spent years tracking the U.S.–China AI rivalry, the erosion of intellectual property protections, and the evolving landscape of digital statecraft. His reporting has broken stories on federal court battles, executive overreach, and the geopolitical dimensions of the global technology race. Follow Samuel and USA Herald on X: @RealUSAHerald