The FBI, NSA and CISA have accused six major Chinese AI companies — including DeepSeek, Alibaba and Moonshot — of systematically extracting capabilities from ChatGPT, Claude, Gemini and Grok. The U.S.-China technology war is moving beyond chips. The next battleground is the intelligence inside the models.
WASHINGTON — The United States has escalated its technology confrontation with China, accusing six leading Chinese artificial-intelligence companies of systematically exploiting American frontier models to accelerate their own development.
In a joint alert, the FBI, National Security Agency and Cybersecurity and Infrastructure Security Agency named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI, alleging that they used outputs from systems including OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini and xAI’s Grok. U.S. officials said the activity had occurred since late 2024 and was likely conducted with Chinese government awareness.
The Dispute Is About More Than “Copying”
At the center of the accusations is distillation — training a smaller model using outputs generated by a more powerful model.
Distillation itself is widely used across the AI industry and is not inherently improper. The American accusation is that Chinese companies conducted it at industrial scale against U.S. systems, violating usage restrictions and extracting capabilities that took American laboratories enormous amounts of capital and computing power to develop.
China has rejected similar accusations as unfounded, arguing that distillation is a standard technical method and accusing Washington of using national-security claims to suppress Chinese competition.
That distinction matters.
The technology is ordinary. The scale, method and geopolitical purpose are what Washington is challenging.
AI Is Becoming a Strategic Asset Like Advanced Chips
The wider significance is that Washington increasingly appears to view frontier AI models themselves as strategic infrastructure.
For years, the U.S.-China technology struggle centered on semiconductors: Nvidia processors, lithography equipment and high-bandwidth memory.
Now the contest is moving upward into the models those chips train.
If American policymakers conclude that unrestricted access to U.S. AI systems allows Chinese competitors to bypass some of the advantages created by chip controls, Washington could tighten API access, identity verification, cloud-computing rules and restrictions on entities believed to be extracting frontier capabilities.
That would represent another stage in technological decoupling.
Awkward Timing for Trump and Xi
The allegations also arrive at a sensitive diplomatic moment.
President Donald Trump and Chinese President Xi Jinping are expected to meet on September 24, while Washington and Beijing are preparing their first formal bilateral discussions focused specifically on advanced-AI safety risks.
That produces an unusual contradiction.
The United States believes cooperation with China may be necessary to prevent catastrophic AI misuse.
At the same time, American security agencies are accusing leading Chinese firms of exploiting U.S. AI laboratories.
Washington and Beijing may therefore be entering AI arms-control discussions while simultaneously accusing one another of technological predation.
WARYATV Assessment
The real battle is no longer simply over who builds the best chatbot.
Artificial intelligence is becoming part of national power — tied to cyberwarfare, intelligence, autonomous weapons, industrial productivity and military planning.
From Washington’s perspective, allowing Chinese companies to cheaply reproduce capabilities developed by American laboratories could erode one of the United States’ remaining technological advantages.
From Beijing’s perspective, American attempts to restrict both advanced chips and access to leading models look increasingly like an effort to preserve technological dominance.
Both interpretations are likely to drive further escalation.
WARYATV Bottom Line
The U.S.-China AI competition has moved from chips to intelligence itself.
Washington is effectively arguing that access to frontier American models is a strategic resource that cannot simply be mined by competitors.
If that principle becomes policy, the next technology controls may not stop at semiconductors.
America may begin building walls around the AI models themselves.
