US Formally Names 6 Chinese AI Firms in Theft Advisory

The NSA, FBI, and CISA jointly issued a cybersecurity advisory accusing six leading Chinese AI companies, DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, of systematically copying American AI models through a technique called distillation, marking the U.S. government's most detailed accusation on this subject to date, according to NBC News's reporting on the advisory.

The report goes into genuinely granular detail, including a long list alleging specific Chinese models distilled particular capabilities, described in the advisory using terms like "legal specialization optimization," "agentic functions," and "coach/assistant capabilities," from specific named American models. The advisory's language was notably blunt about the scale involved: "The sheer scale of these campaigns and their sophistication indicate that distillation is not a supplement to these companies' AI model development, but the critical core of it," according to NBC News's reporting.

The advisory identified American models from Anthropic, OpenAI, Google, and SpaceX as the specific targets of this activity, according to Parameter's reporting on the disclosure. This isn't the U.S. government's first accusation of this kind, but it is the first time these specific allegations have been formalized jointly across three major federal agencies, rather than surfacing as isolated corporate complaints. American companies have made scattered accusations of distillation since at least last year, when OpenAI said DeepSeek had "inappropriately" based its model on ChatGPT, and Google separately said in February it was inundated with attempts to clone Gemini but declined to name a specific suspect at the time.

The US-China AI Distillation Advisory at a Glance

Distillation is a widely used, legitimate AI training technique where developers train smaller, cheaper models using outputs generated by larger, more expensive systems, a practice that reduces development costs and is common across the AI industry broadly, according to MarketScreener's reporting on the technique. U.S. officials described the specific practices attributed to these six companies as "aggressive, malicious and targeted," a characterization that distinguishes what they're alleging from the routine, permitted use of distillation elsewhere in the industry. This tension connects directly to our own earlier coverage of this exact controversy, including NBC News's reporting on Kimi K3 rivaling Claude and ChatGPT amid the same underlying distillation dispute and Treasury Secretary Scott Bessent's earlier warning that the US could sanction China over alleged AI intellectual property theft.

This advisory's timing lands at a genuinely sensitive diplomatic moment. The accusations arrive just weeks ahead of a scheduled September 24 meeting between President Trump and President Xi Jinping, during which AI safety and technological competition are expected to be central topics, according to Time News's reporting on the advisory. The U.S. and China are also separately preparing for a mid-September dialogue specifically concerning AI safety risks, a meeting we've tracked closely alongside Washington telling dozens of allies they must pick sides in the AI race with China earlier this summer.

The response from the named companies has been largely muted. Moonshot AI declined to address the claims directly, while the other named firms did not immediately respond to requests for comment, according to Time News's reporting. China has separately dismissed the U.S. claims broadly, though the Chinese embassy has not issued detailed public comment specifically addressing this advisory, according to Ground News' reporting on the dispute.

This advisory is worth understanding for any business currently using or evaluating open-weight AI models from any of these six named companies, given the formalized nature of this accusation, now backed jointly by three major U.S. intelligence and security agencies rather than isolated corporate complaints, represents a genuine escalation in official government scrutiny that could eventually translate into export controls, sanctions, or other trade restrictions.

For businesses evaluating AI vendor risk more broadly, this advisory reinforces the value of building genuine model flexibility into technical architecture now, since the geopolitical status of specific Chinese AI providers could shift meaningfully and quickly depending on how the upcoming Trump-Xi meeting and broader U.S.-China AI dialogue unfold.

Which Chinese AI companies did the US formally accuse of copying American AI?
The NSA, FBI, and CISA jointly named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI in a cybersecurity advisory accusing them of systematically distilling capabilities from American AI models.

What is AI distillation, and why is it controversial in this case?
Distillation is a legitimate technique where smaller AI models are trained using outputs from larger, more expensive models, commonly used to reduce development costs. U.S. officials allege the practices attributed to these six companies went beyond legitimate use, describing them as "aggressive, malicious and targeted."

Why is the timing of this advisory significant?
The advisory arrives just weeks before a scheduled September 24 meeting between President Trump and President Xi Jinping, where AI safety and technological competition are expected to be central discussion topics, alongside a separately planned U.S.-China AI safety dialogue.

The NSA, FBI, and CISA jointly issued a cybersecurity advisory naming six Chinese AI companies, DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, accusing them of systematically distilling capabilities from American AI models developed by Anthropic, OpenAI, Google, and SpaceX since late 2024. The advisory describes the alleged activity as "aggressive, malicious and targeted," distinguishing it from the routine, legitimate use of distillation elsewhere in the AI industry, and states the practices likely occurred with Chinese government awareness. The accusations land just weeks before a scheduled September 24 meeting between President Trump and President Xi Jinping, where AI competition is expected to be a central discussion topic, with most of the named companies declining to respond directly to the claims.