AI War? White House accuses China of industrial-scale theft of US AI frontier models

The White House on Thursday accused China-based actors of targeting U.S. artificial intelligence labs in coordinated campaigns to extract intellectual property.

The warning comes just weeks before U.S. President Donald Trump is expected to meet Chinese President Xi Jinping in Beijing, raising concerns that the issue could overshadow upcoming talks.

Industrial-scale AI extraction

Michael Kratsios, director of the White House Office of Science and Technology Policy, detailed the allegations in a memo circulated to federal agencies.

“The United States government has information indicating that foreign entities, principally based in China, are engaged in deliberate, industrial-scale campaigns to distill US frontier AI systems,” he wrote in a memo.

Kratsios said the campaigns rely on scale and coordination. He noted that actors use proxy accounts to avoid detection while probing AI systems.

He also said they use jailbreaking techniques to expose restricted or proprietary outputs.

Officials believe these operations involve querying advanced AI systems repeatedly.

Attackers analyze outputs to understand how models behave across different prompts.

Over time, this allows them to reconstruct capabilities without direct access to the underlying systems.

Distillation tactics explained

The memo centers on a technique known as distillation.

This process involves training smaller AI models using outputs generated by larger, more advanced systems.

By sending millions of queries through application programming interfaces, attackers can collect large datasets.

These datasets help recreate how a model responds, enabling the development of competing systems at lower cost.

Kratsios warned that such campaigns can extract core capabilities from American models.

He added that they exploit U.S. innovation while bypassing the time and expense required to build frontier systems.

The memo also raises concerns about safeguards.

Officials say distillation efforts can weaken guardrails designed to ensure outputs remain accurate and controlled.

The Chinese Embassy in Washington pushed back strongly against the claims.

It said it opposes “the baseless allegations,” and added that Beijing “attaches great importance to the protection of intellectual property rights” as reported by Reuters.

Rising tech war tensions

The accusations could deepen an already tense technology relationship between the United States and China.

They also introduce uncertainty around key policy decisions tied to AI and semiconductors.

One major question involves exports of advanced chips from NVIDIA.

The administration approved conditional sales earlier this year, but officials have indicated that shipments have not yet begun.

U.S. AI companies have raised similar concerns in recent months.

Firms including OpenAI and Anthropic reported large-scale distillation attempts targeting their systems.

They linked these efforts to Chinese firms such as DeepSeek, Moonshot AI, and MiniMax.

The U.S. has also pursued legal action tied to AI theft. In 2024, prosecutors charged a former engineer at Google with stealing AI trade secrets and sharing them with Chinese companies.

Kratsios also cast doubt on the durability of such copied systems.

He said models built on extracted capabilities may struggle as detection methods improve.

The administration said it plans to share intelligence with American AI firms and explore measures to hold those responsible accountable.

The memo signals a more confrontational stage in the U.S.-China AI rivalry, with advanced AI systems now at the center of geopolitical competition.