America Says China Is Copying Its AI Homework — at Industrial Scale

🎥 Video Overview

📖 Explanation (Ages 14–18)

What if someone could clone your most advanced technology not by stealing blueprints, but by asking your own product clever questions until it spilled its secrets?

📖 What's Going On?

The White House has formally accused Chinese entities of running massive, coordinated campaigns to steal the capabilities of America's most advanced AI systems. Michael Kratsios, the director of the White House Office of Science and Technology Policy, issued a memo warning that Chinese groups are using a technique called 'distillation' — essentially training cheaper, smaller AI models by feeding them the outputs of America's frontier models — at an unprecedented scale.

According to the memo, these campaigns involve tens of thousands of fake proxy accounts to dodge detection and 'jailbreaking' techniques that trick AI systems into revealing proprietary information they're designed to protect. The accusation lands at a politically charged moment: President Trump is set to meet President Xi Jinping in Beijing within weeks. China's embassy called the claims 'pure slander,' insisting the country respects intellectual property rights and promotes cooperation.

🎯 How To Think About It

Distillation is a legitimate technique in AI — researchers use it all the time to compress a powerful model into a lighter, cheaper version. The controversy is about who's doing it, how, and whether they have permission. Think of two parallels:

• Imagine a sneaker reseller reverse-engineering Nike's proprietary foam formula by buying hundreds of pairs and chemically analyzing them — they didn't break into the factory, but they're still extracting trade secrets from the product itself. That's roughly what unauthorized distillation does to an AI model.
• It's also like the Cold War technology race: the Soviet Union famously copied the American B-29 bomber bolt-by-bolt after capturing one. The US had export controls then too, but physical objects are easier to guard than digital outputs flowing through API endpoints across the internet.

💡 Key Things To Know

• In February 2026, Anthropic accused three Chinese AI firms — DeepSeek, Moonshot, and MiniMax — of conducting distillation attacks on its models. OpenAI had flagged similar concerns about DeepSeek as early as 2025.
• Distillation works because a smaller model can learn patterns from a larger model's outputs without needing the same massive computing infrastructure. This lets Chinese firms sidestep US chip export controls that were designed to slow their progress.
• The House Foreign Affairs Committee passed new bills this week that would add companies caught distilling to the 'entity list' — an export blacklist that effectively cuts them off from American technology supply chains.
• Distilled models often lack the safety guardrails built into the originals, raising national security concerns about misuse for bioweapons development or cyberattacks.
• Most people assume AI theft means hacking into servers and stealing code. The counterintuitive part is that distillation doesn't require any break-in — attackers extract value by systematically querying the model through its normal interface, making it far harder to detect and prosecute.

🌟 Why It Matters

If you're considering a career in tech, AI research, cybersecurity, or international policy, this is the defining conflict of the field right now. The outcome will shape which countries control the most powerful AI systems, how open or locked-down those systems become, and whether the research culture in AI stays collaborative or fractures along national lines. It also affects everyday users: tighter security could mean more restricted access to AI tools, higher costs, and slower innovation as companies prioritize defense over openness.

🔮 The Bigger Picture

The US-China AI rivalry is rapidly becoming the 21st century's central technological competition, much as the nuclear and space races defined the Cold War. If the US cracks down hard — banning Chinese access to models, sanctioning distillation, tightening chip exports — it could fragment the global AI ecosystem into two separate blocs. Watch for second-order effects: American AI companies may restrict API access for all international users, open-source AI development could face new legal scrutiny, and allies in Europe and Asia will be pressured to pick sides. The upcoming Trump-Xi summit will be a critical signal of whether this escalates into a full technological decoupling or whether some diplomatic middle ground emerges.

📚 Key Terms Glossary