AI researchers are publicly questioning a claim made by White House science advisor Michael Kratsios that the Chinese company Moonshot built its Kimi K3 AI model by copying Anthropic's Fable model, using computer chips that the U.S. had not approved for export to China. On July 23, 2026, TechCrunch reported that multiple experts consider the copying claim technically implausible given the timelines involved (TechCrunch).

Kratsios described the alleged copying as "Large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research." He did not share details about the sources of his allegations. Moonshot did not respond to TechCrunch's questions about its training process.

The technique Kratsios referred to is called distillation. In AI, distillation means using a powerful existing model's answers to train a new model. Think of it like a student learning by reading a teacher's exam answers rather than working through the textbook. It is a known method, but the question here is whether it could have produced Kimi K3 in the time available.

The claims come amid reported discussions within the U.S. government about banning Chinese AI models that are released as "open-weight," meaning their underlying code is published for anyone to download and use. Treasury Secretary Scott Bessent said "we are finding watermarks of our U.S. large language models on many of the Chinese models, and that that's unacceptable." The Treasury Department did not respond to TechCrunch's query about what those watermarks consist of. Treasury had separately threatened sanctions after the White House claims regarding Moonshot's distillation of Fable (TechCrunch).

The technical pushback from researchers centers on timing and feasibility. Braden Hancock, a researcher at the Laude Institute and co-founder of Snorkel AI, told TechCrunch he does not think Kimi K3's strength came from copying Anthropic's Fable. Hancock noted that Fable was only publicly available since July 1, 2026, leaving insufficient time to extract training data, train a model, and release it within two weeks. Nathan Lambert, an AI researcher at the Allen Institute for AI, said distillation has become less impactful over time as Chinese models get closer to the frontier and training shifts to reinforcement learning, a method where models learn through trial and error rather than imitation. Lambert argued that if copying were the key factor, competitors could catch up to GLM or K3 using their data the same way, but that has not happened from imitation alone. He also said that copying Fable's capabilities would likely require reinforcement learning with tens of millions of agents, which would be prohibitively expensive and a time bottleneck when using a frontier lab's API.

The broader dispute builds on a longer-running conflict. Anthropic publicly accused Moonshot, DeepSeek, and MiniMax of systematically copying its models earlier in 2026 (Anthropic). In a February 2026 post, Anthropic warned that if copied models are released publicly, risk multiplies as capabilities spread freely beyond any single government's control. A subsequent June 2026 Anthropic post on Claude Fable 5 and Claude Mythos 5 stated that copying Fable 5's abilities could indirectly lead to the spread of near-frontier AI capabilities (Anthropic). Anthropic's research publication "2028: Two scenarios for global AI leadership" discussed how Chinese labs have remained close to the frontier by exploiting U.S. export control loopholes and carrying out large-scale copying (Anthropic).

Kimi K3 was released on July 16, 2026. The model is natively multimodal, meaning it can process text, images, and other types of input together. It features a 1-million-token context window, meaning it can consider a very large amount of text in a single conversation, and is designed for long-horizon coding and end-to-end knowledge work (Moonshot AI). The Kimi K2 series was officially discontinued on May 25, 2026. Moonshot has stated that Kimi K3 "still trails the most powerful proprietary models, Claude Fable 5 and GPT 5.6 Sol" (TechCrunch.

Moonshot stated that full model weights for K3 will be released by July 27, 2026, and that the company will publish more details on the model's architecture, training methodology, and evaluation (Moonshot AI).

The broader context here is that this dispute is not just a technical argument. The U.S. government is weighing restrictions on Chinese open-weight models, and the Kratsios allegations surfaced in the context of those discussions. The technical merits of the copying claim matter because they underpin a proposed regulatory action that could restrict access to open-weight models broadly, not just those from Moonshot. If the copying pathway described by Kratsios is not feasible within the available timeframe, as Hancock and Lambert argue, the evidentiary basis for a ban weakens considerably.

What also remains unclear is what Bessent meant by "watermarks" found on Chinese models. Without a technical definition from Treasury, the term could refer to anything from residual traces in model outputs to markers deliberately embedded in training data. The distinction matters. A deliberate watermark embedded by Anthropic and subsequently detected would be materially different from indirect behavioral patterns that could arise from different labs training on overlapping data.

The public release of Kimi K3 on July 27 may shed additional light. If Moonshot follows through on its commitment to publish training methodology details, independent researchers will have an opportunity to evaluate the copying claims directly. Until then, the allegations rest on government assertions that have not been substantiated with technical evidence, set against expert skepticism from researchers with direct experience in distillation techniques.

In my view, the gap between the political framing and the technical reality is the most consequential part of this story. We have seen this pattern before, when regulatory urgency outpaces the evidence needed to justify it. When allegations of this magnitude are made without accompanying technical detail, it becomes harder for the public to distinguish legitimate national security concerns from competitive maneuvering between rival labs and governments.