美国官员指控Moonshot蒸馏Anthropic并通过泰国访问GB300，先进GPU跨境合规风险升温

# 美股盘后催化 9/11: 美国威胁制裁中国AI模型蒸馏并指控Moonshot涉GB300访问，先进GPU跨境合规风险升温

- 美股会话: 2026-07-22
- 方向: MIXED；类型: 监管/政策；规划重大性: 8/10
- 映射模式: competitive
- 美股驱动（规划器初判，需独立核验）: 事实：白宫官员指控Moonshot通过大规模蒸馏Anthropic模型开发Kimi K3，并可能通过泰国访问NVIDIA GB300服务器；美国财政部长称从事隐蔽工业级蒸馏攻击的中国公司可能面临制裁和实体清单。反向材料显示美国开源AI从业者和NVIDIA CEO均淡化中国开源模型威胁。推断：对中国大模型、跨境算力租赁和GPU转运链是政策负面，对国产算力和合规云替代是中长期正面，对NVDA短期是需求强但出口合规风险上升。

## 本催化相关美股真实收盘行情

- 英伟达 (NVDA.US): 212.06 +2.30%
- 微软 (MSFT.US): 390.34 -1.86%
- 谷歌 (GOOGL.US): 342.09 -1.46%
- Meta (META.US): 627.17 -2.58%
- 亚马逊 (AMZN.US): 244.85 -1.09%

## 亚太映射假设（仅候选，必须查证）

- 阿里巴巴 (9988.HK) [HK/竞争替代/同源本地股/mixed]: 中国大模型和云生态可能受美国审查外溢压制，但国内自主可控需求可能提升。；核验问题: 核实阿里云/通义是否被纳入类似审查或客户采购限制。
- 百度集团 (9888.HK) [HK/竞争替代/同源本地股/mixed]: 文心及百度智能云海外部署可能受到模型来源和算力合规叙事影响。；核验问题: 核实百度模型训练数据、海外客户和芯片获取路径的监管暴露。
- 寒武纪 (688256.SH) [CN/竞争替代/positive]: 若高端海外GPU访问受限，国产AI加速器采用意愿可能提升。；核验问题: 核实国内大模型厂商是否扩大寒武纪等国产加速器采购和适配。
- 中芯国际 (0981.HK) [HK/供应商/positive]: 国产AI芯片替代需求提升可能增加本土制造需求，但先进制程能力限制需核实。；核验问题: 核实相关国产AI芯片是否由中芯国际制造及封装可行性。

## 必答问题

- 美国是否会正式把Moonshot或相关云/分销商列入实体清单？
- GB300访问指控是否有可验证证据和NVIDIA渠道责任？
- 中国AI公司是否会加速国产芯片迁移并披露采购计划？

## 原始证据全文（13 条）

### [E1205] Bloomberg | 2026-07-22 15:45 UTC | 盘中
White House Official Says Moonshot Accessed Banned Nvidia Chips
URL: https://www.bloomberg.com/news/articles/2026-07-22/white-house-official-says-moonshot-accessed-banned-nvidia-chips
A White House official accused China’s Moonshot of improperly using US AI models and Nvidia Corp. chips to create the Kimi K3 system that stunned the tech industry last week with its advanced capabilities.

The company “acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models,” White House Office of Science and Technology Policy Director Michael Kratsios said in a social media post Wednesday, making the case that Moonshot violated US export control rules and companies’ terms of service to build its product.

The semiconductor servers Kratsios referenced are part of the Blackwell generation of Nvidia products, which the US forbids from sale to Chinese companies. Though the Trump administration has allowed some sales of Nvidia’s H200 chips, it continues to prevent sales of the more advanced Blackwells.

A May memo affirmed that the Bureau of Industry and Security, which is responsible for export controls on AI chips, restricts sale of advanced processors to Chinese companies outside of China. Trump officials are now looking into whether some companies took advantage of a loophole to ship advanced AI chips to subsidiaries of Chinese firms.

An Nvidia spokesperson earlier dismissed the idea of a loophole and said it has operated as though it does not exist. The company did not respond to a request for comment on the Kratsios post.

Moonshot’s K3 upended assumptions that Chinese AI capabilities were meaningfully behind US leaders such as OpenAI and Anthropic PBC. The Chinese lab said its open-weight model outperforms all rivals except for Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 on overall capability, setting off a rout in tech stocks similar to DeepSeek’s impact last year.

Moonshot “developed a sophisticated internal platform to conduct large scale distillation against US models,” Kratsios said, referencing a technique in which AI researchers use the outputs of a so-called teacher model to train a student model and replicate some of its capabilities.

Treasury Secretary Scott Bessent on Tuesday voiced similar distillation concerns, though not specific to Moonshot, about open-source Chinese AI companies engaging in what he called intellectual property theft to create AI models.

“This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them,” Bessent said on Fox Business.

Read More: Bessent Says US to Scrutinize Chinese AI Models for IP Theft

Bessent’s comments echoed growing accusations from OpenAI and Anthropic that their Chinese rivals have made systematic and unauthorized use of results from leading US models to develop a rival generation of chatbots at a fraction of the cost. Anthropic has previously accused Mooonshot specifically of distilling its models. Moonshot has not responded to those allegations.

In April, the Trump administration vowed to take further steps to address the practice, known as adversarial distillation, including weighing steps to hold foreign actors accountable. Lawmakers in the House and the Senate have advanced legislation with similar aims of blacklisting or sanctioning Chinese firms found to be improperly distilling from US models.

### [E1206] Bloomberg | 2026-07-22 15:45 UTC | 盘中
White House Official Says Moonshot Accessed Banned Nvidia Chips
URL: https://www.bloomberg.com/news/articles/2026-07-22/white-house-official-says-moonshot-accessed-banned-nvidia-chips
A White House official accused China’s Moonshot of improperly using US AI models and Nvidia Corp. chips to create the Kimi K3 system that stunned the tech industry last week with its advanced capabilities.

The company “acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models,” White House Office of Science and Technology Policy Director Michael Kratsios said in a social media post Wednesday, making the case that Moonshot violated US export control rules and companies’ terms of service to build its product.

The semiconductor servers Kratsios referenced are part of the Blackwell generation of Nvidia products, which the US forbids from sale to Chinese companies. Though the Trump administration has allowed some sales of Nvidia’s H200 chips, it continues to prevent sales of the more advanced Blackwells.

A May memo affirmed that the Bureau of Industry and Security, which is responsible for export controls on AI chips, restricts sale of advanced processors to Chinese companies outside of China. Trump officials are now looking into whether some companies took advantage of a loophole to ship advanced AI chips to subsidiaries of Chinese firms.

An Nvidia spokesperson earlier dismissed the idea of a loophole and said it has operated as though it does not exist. The company did not respond to a request for comment on the Kratsios post.

Moonshot’s K3 upended assumptions that Chinese AI capabilities were meaningfully behind US leaders such as OpenAI and Anthropic PBC. The Chinese lab said its open-weight model outperforms all rivals except for Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 on overall capability, setting off a rout in tech stocks similar to DeepSeek’s impact last year.

Moonshot “developed a sophisticated internal platform to conduct large scale distillation against US models,” Kratsios said, referencing a technique in which AI researchers use the outputs of a so-called teacher model to train a student model and replicate some of its capabilities.

Treasury Secretary Scott Bessent on Tuesday voiced similar distillation concerns, though not specific to Moonshot, about open-source Chinese AI companies engaging in what he called intellectual property theft to create AI models.

“This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them,” Bessent said on Fox Business.

Read More: Bessent Says US to Scrutinize Chinese AI Models for IP Theft

Bessent’s comments echoed growing accusations from OpenAI and Anthropic that their Chinese rivals have made systematic and unauthorized use of results from leading US models to develop a rival generation of chatbots at a fraction of the cost. Anthropic has previously accused Mooonshot specifically of distilling its models. Moonshot has not responded to those allegations.

In April, the Trump administration vowed to take further steps to address the practice, known as adversarial distillation, including weighing steps to hold foreign actors accountable. Lawmakers in the House and the Senate have advanced legislation with similar aims of blacklisting or sanctioning Chinese firms found to be improperly distilling from US models.

### [E1301] Seeking Alpha | 2026-07-22 16:00 UTC | 盘中
Trump official says Moonshot built Kimi K3 through theft of Anthropic's Fable
URL: https://seekingalpha.com/news/4616700-kratsios-says-moonshot-built-kimi-k3-through-industrial-distillation-of-anthropics-fable?utm_source=feed_news_all&utm_medium=referral&feed_item_type=news
The China-based startup Moonshot AI developed its latest mode, Kimi K3, through the "industrial distillation" of Anthropic's (ANTHRO) Fable, according to Michael Kratsios, the science advisor to U.S. President Donald Trump.
"We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," Kratsios said in a post on X Wednesday morning.
What's more, Moonshot also gained access to Nvidia's (NVDA) GB300 through servers located in Thailand to help train its models. Under U.S. Department of Commerce rules, Nvidia's GB300 chips are not allowed for export to China.
"To do this, they developed a sophisticated internal platform to conduct large-scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection," Kratsios said. "Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models."
On Tuesday, U.S. Treasury Secretary Scott Bessent said the Trump Administration is entitled to sanction open-weight models from China if evidence of distillation from American models is verified.
"The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models," Kratsios added. "Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable."
Distillation is a training method where less capable models are improved by utilizing outputs from stronger models.
Kimi K3 caused a stir through the AI market when it was released last week. Moonshot said the 2.8 trillion-parameter system is the world's largest open-weight AI model. Kimi K3 "performed competitively" with Fable 5 and "substantially outperformed" OpenAI's (OPENAI) Opus 4.8, GPT 5.6 Sol and GPT 5.5 in terms of GPU kernel optimization, the company said.
Dear Readers: We recognize that politics often intersect with the financial news of the day, so we invite you to click here to join the separate political discussion.

### [E1312] Investing.com | 2026-07-22 16:03 UTC | 盘中
White House accuses Moonshot of violating US export controls using Nvidia hardware
URL: https://www.investing.com/news/stock-market-news/white-house-accuses-moonshot-of-violating-us-export-controls-using-nvidia-hardware-4806584
Wall Street mixed as investors eye tech earnings for AI cues, led by Alphabet
Investing.com -- A top White House official accused Chinese AI startup Moonshot of illicitly accessing cutting-edge US technology, alleging the lab used NVIDIA hardware and proprietary American AI models to build its flagship Kimi K3 system.
Michael Kratsios, Director of the White House Office of Science and Technology Policy, stated on social media that Moonshot acquired servers equipped with NVIDIA GB300 chips and routed workloads through GB300-enabled infrastructure in Thailand to train its models. Kratsios argued that these actions violate both US export regulations and the terms of service set by American AI developers.
The semiconductor systems in question belong to NVIDIA’s Blackwell generation (specifically the GB300 platform), which the US government explicitly bars from export to Chinese companies. While the US administration has granted exemptions for lower-tier processors such as the H200, it maintains a strict ban on high-end Blackwell hardware.
Federal investigators at the Bureau of Industry and Security (BIS)—the agency enforcing technology transfer restrictions—are examining whether Chinese entities have bypassed restrictions by utilizing overseas subsidiaries or cloud infrastructure in third-party countries like Thailand.
NVIDIA has previously maintained that it complies with all export control regulations and actively enforces compliance across its sales channels. The company did not immediately comment on the White House’s latest statement.
The dispute comes on the heels of Moonshot’s Kimi K3 release, an open-weight model that surprised industry analysts by narrowing the performance gap between Chinese labs and US frontier AI companies. Moonshot claims K3 rivals leading American models, asserting it trails only OpenAI’s GPT-5.6 and Anthropic’s Claude Fable 5 in benchmark testing.
According to Kratsios, Moonshot achieved these gains in part by developing an internal platform to conduct model distillation at scale. Distillation is a technique where developers use the outputs of advanced "teacher" models—in this case, proprietary US systems—to rapidly train smaller or open-weight "student" models to replicate their capabilities.

### [E1443] Tech Insider | 2026-07-22 16:34 UTC | 盘中
A top White House official is escalating the fight over Moonshot AI's viral Kimi K3 model
URL: https://www.businessinsider.com/white-house-kimi-k3-moonshot-ai-distillation-2026-7
A top White House official publicly accused Moonshot AI of essentially stealing from Anthropic, intensifying the debate over China-based open source AI models.
"We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," Michael Kratsios, director of the White House Office of Science and Technology Policy, wrote on X Wednesday morning.
We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model.
— Director Michael Kratsios (@mkratsios47) July 22, 2026
To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of…
Kratsios said Moonshot AI, the maker of the viral Kimi K3 model, went to great lengths to conceal its actions.
"To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection," he wrote.
Kimi K3 is widely regarded as the most impressive open source AI model released thus far. Moonshot's model has performed at or above the level of Anthropic and OpenAI's leading frontier models in benchmark testing. It is likely that both US companies spent billions developing those models. In contrast, Moonshot is expected to make Kimi K3 freely available for download on local devices and customizable to a user's content when it releases the model's full weights next week.
Distillation by itself is not wrong, a point that Kratsios stressed in his statement. AI labs themselves use the process of training a less powerful model on the output of a more powerful model before releasing updates.
"Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem," he wrote. "However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable."
Kratsios' statement comes after Treasury Secretary Scott Bessent suggested that the US government may sanction Chinese companies if it is proven that their models were improperly trained through a process called distillation.
The White House's words so far are notable, as they mark a rare instance in which the Trump administration does not find itself at loggerheads with Anthropic. The Trump administration previously imposed export controls to block Anthropic from releasing Fable 5, its most advanced model, and in February the Pentagon moved to blacklist the AI giant.
"Thanks @mkratsios47 for speaking out on this important issue," Sarah Heck, Anthropic's head of public policy, wrote on X, quoting Kratsios' statement.
Heck said Anthropic will continue to work with the White House and Congress on the issue.
"Illicit, adversarial distillation is IP theft and industrial espionage that supports adversary military and intelligence capabilities," she wrote. "It is a national challenge that creates serious national security risks for the United States and democratic allies."
Increasingly, US companies have said that their Chinese competitors have abused the process to rip off their intellectual property. In February, Anthropic publicly accused three companies, including Moonshot AI, of using distillation attacks that violate its terms of service.
A spokesperson for Anthropic did not immediately respond to a request for comment from Business Insider.

### [E1523] Reuters Technology | 2026-07-22 16:55 UTC | 盘中
US accuses China's Moonshot of stealing from Anthropic's Fable for latest AI model
URL: https://www.reuters.com/world/us/trump-reports-more-than-14-billion-income-crypto-ventures-2026-06-30/
China's Moonshot ​AI stole from Anthropic's most sophisticated AI large language model, Fable, to help create its latest ‌K3 release and has acquired advanced Nvidia AI chips, top White House tech official Michael Kratsios said on Wednesday.

"We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model," he said in a post on ​X. "Large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is ​unacceptable."

He also said Moonshot AI had acquired servers containing powerful Nvidia (NVDA.O) AI chips called ⁠GB300s and had used them in Thailand, "likely to train its AI models." A senior Trump administration ​official said in February that Chinese AI startup DeepSeek's latest AI model was trained on Nvidia's most advanced AI chip, ​in what could represent a violation of U.S. export controls.

Moonshot and the Chinese Embassy in Washington did not immediately respond to a request for comment.

The U.S. State Department ordered a global push in April to bring attention to what it ​says are widespread efforts by Chinese companies, including Moonshot AI, to steal intellectual property from U.S. ​artificial intelligence labs, according to a diplomatic cable seen by Reuters.

Distillation is the process of training smaller AI models using ‌output ⁠from larger, more expensive ones as part of an effort to lower the costs of training a powerful new AI tool.

OpenAI has warned U.S. lawmakers that DeepSeek was targeting the ChatGPT maker and the nation's leading AI companies to replicate models and use them for its own training, The ​April cable from the State ​Department said, "AI models ⁠developed from surreptitious, unauthorized distillation campaigns enable foreign actors to release products that appear to perform comparably on select benchmarks at a fraction of the cost ​but do not replicate the full performance of the original system."

Chinese AI ​startup Moonshot on ⁠Friday unveiled Kimi K3, a 2.8 trillion-parameter model that it said is the world's largest open-weight AI system and delivers performance approaching U.S. giant Anthropic's frontier Fable model.

The launch, which comes a month after Anthropic's ⁠Fable and ​Mythos models were abruptly withdrawn by the U.S. government due ​to security concerns, underscores how quickly China's open AI ecosystem is narrowing the gap with the most advanced U.S. systems.

### [E1707] First Squawk | 2026-07-22 17:45 UTC | 盘中
WHITE HOUSE ACCUSES CHINESE AI STARTUP MOONSHOT OF ILLEGALLY USING BANNED NVIDIA BLACKWELL CHIPS AND DISTILLING U.S. AI MODELS TO DEVELOP ITS KIMI K3 MODEL; OFFICIALS SAY RESTRICTED CHIPS MAY HAVE BEEN ACCESSED THROUGH THAILAND IN VIOLATION
URL: https://x.com/FirstSquawk/status/2079986262577152256
WHITE HOUSE ACCUSES CHINESE AI STARTUP MOONSHOT OF ILLEGALLY USING BANNED NVIDIA BLACKWELL CHIPS AND DISTILLING U.S. AI MODELS TO DEVELOP ITS KIMI K3 MODEL; OFFICIALS SAY RESTRICTED CHIPS MAY HAVE BEEN ACCESSED THROUGH THAILAND IN VIOLATION OF U.S. EXPORT CONTROLS

MOONSHOT HAS NOT RESPONDED TO ALLEGATIONS; NVIDIA STATES IT DOES NOT EXPLOIT EXPORT-CONTROL LOOPHOLES

### [E2046] First Squawk | 2026-07-22 19:17 UTC | 盘中
U.S. TREASURY SECRETARY BESSENT: CHINESE FIRMS ENGAGING IN COVERT INDUSTRIAL-SCALE DISTILLATION ATTACKS FACE POTENTIAL SANCTIONS AND ENTITY LIST DESIGNATIONS
URL: https://x.com/FirstSquawk/status/2080009413063270595
U.S. TREASURY SECRETARY BESSENT: CHINESE FIRMS ENGAGING IN COVERT INDUSTRIAL-SCALE DISTILLATION ATTACKS FACE POTENTIAL SANCTIONS AND ENTITY LIST DESIGNATIONS

### [E2681] TechCrunch AI | 2026-07-22 20:49 UTC | 收盘后
Treasury threatens sanctions after White House claims Moonshot distilled Anthropic’s Fable
URL: https://techcrunch.com/2026/07/22/treasury-threatens-sanctions-after-white-house-claims-moonshot-distilled-anthropics-fable/
U.S. Treasury Secretary Scott Bessent doubled down on his warnings to Chinese AI companies on Wednesday, saying that sanctions remain on the table after a White House official accused Moonshot of improperly distilling Anthropic’s Fable model.
Model distillation is a common AI training technique in which a smaller model learns from the outputs of a larger one. While this process can infringe on intellectual property rights, it’s also widely used as a legitimate optimization method.
“Open source is not open season on American IP,” Bessent posted on X. “When [Chinese] firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.”
Earlier this week, Bessent stated that the U.S. government would examine open source models from China for signs of intellectual property theft and impose sanctions if found.
Bessent’s latest remarks come hours after the White House’s science and technology policy chief Michael Kratsios accused the China-based Moonshot of conducting large scale distillation against U.S. models. He alleged that Moonshot had acquired Nvidia’s “GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models,” raising questions about whether the firm violated U.S. export control rules.
The GB300 servers are part of Nvidia’s Blackwell generation, which are banned from being sold to Chinese companies.
Some experts dispute the idea that Kimi K3 could have been developed primarily through distillation from Fable, which has only been publicly available since July 1. Moonshot released K3 last week as an open-weight model, and its advanced capabilities have called into question the underlying business models of leading U.S. AI labs, casting doubt on whether they can continue to justify the enormous capital requirements underpinning the frontier AI race.
The episode has also intensified a broader debate in Washington over the influx of Chinese open models. Some, including former White House AI advisor and current OpenAI Head of Strategic Futures, Dean Ball, have argued that the U.S. should restrict or effectively ban the use of Chinese open-weight models to preserve America’s technological advantage and mitigate potential national security risks.
TechCrunch has reached out to Moonshot and the Treasury for comment.

### [E2682] TechCrunch AI | 2026-07-22 20:49 UTC | 收盘后
Treasury threatens sanctions after White House claims Moonshot distilled Anthropic’s Fable
URL: https://techcrunch.com/2026/07/22/treasury-threatens-sanctions-after-white-house-claims-moonshot-distilled-anthropics-fable/
U.S. Treasury Secretary Scott Bessent doubled down on his warnings to Chinese AI companies on Wednesday, saying that sanctions remain on the table after a White House official accused Moonshot of improperly distilling Anthropic’s Fable model.
Model distillation is a common AI training technique in which a smaller model learns from the outputs of a larger one. While this process can infringe on intellectual property rights, it’s also widely used as a legitimate optimization method.
“Open source is not open season on American IP,” Bessent posted on X. “When [Chinese] firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.”
Earlier this week, Bessent stated that the U.S. government would examine open source models from China for signs of intellectual property theft and impose sanctions if found.
Bessent’s latest remarks come hours after the White House’s science and technology policy chief Michael Kratsios accused the China-based Moonshot of conducting large scale distillation against U.S. models. He alleged that Moonshot had acquired Nvidia’s “GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models,” raising questions about whether the firm violated U.S. export control rules.
The GB300 servers are part of Nvidia’s Blackwell generation, which are banned from being sold to Chinese companies.
Some experts dispute the idea that Kimi K3 could have been developed primarily through distillation from Fable, which has only been publicly available since July 1. Moonshot released K3 last week as an open-weight model, and its advanced capabilities have called into question the underlying business models of leading U.S. AI labs, casting doubt on whether they can continue to justify the enormous capital requirements underpinning the frontier AI race.
The episode has also intensified a broader debate in Washington over the influx of Chinese open models. Some, including former White House AI advisor and current OpenAI Head of Strategic Futures, Dean Ball, have argued that the U.S. should restrict or effectively ban the use of Chinese open-weight models to preserve America’s technological advantage and mitigate potential national security risks.
TechCrunch has reached out to Moonshot and the Treasury for comment.

### [E1397] TechCrunch AI | 2026-07-22 16:24 UTC | 盘中
Arcee, a US open source AI lab, says Chinese models are not inherently dangerous
URL: https://techcrunch.com/2026/07/22/arcee-a-us-open-source-ai-lab-says-chinese-models-are-not-inherently-dangerous/
As Chinese open-weight AI models grow in capability and popularity, arguments about what should be done about them have once again reached a fever pitch.
There’s talk that the Trump administration might try to ban them (though it hasn’t yet acted on the idea). Meanwhile, proprietary model makers, particularly OpenAI and Anthropic, appear increasingly concerned about them.
Open-weight models such as Moonshot AI’s Kimi K3 or Alibaba’s Qwen offer inference at a fraction of the token cost of closed-source models from these large U.S. labs. The fear is that they also pose some sort of threat. Certainly, they threaten the profit margins of the large proprietary AI labs.
But should enterprises running these models in their own data centers succumb to the fear that they could be a vector for Chinese hackers?
No, says Lucas Atkins, the CTO of Arcee, which is building open models to give U.S. companies a homegrown alternative to Chinese models.
If any startup would benefit from a ban on Chinese models, Arcee would. But Atkins says China’s open models are no more dangerous than any other open-source software a company may use. In fact, he says, they even offer benefits even to his own company.
“A lot of people view this as similar to a Chinese software program. Like, it was coded with these x, y, z intentions” that a bad actor could simply command, he said.
“That is fundamentally not how these models are trained. There is really not any way for an Arcee, or an Alibaba, to make a model, have someone run it in their own environment and for us have any access to it whatsoever,” he explained.
While most of these models are what’s known as “open weight,” and are not really fully open-source software, the source code (the part that will actually run on servers), if it is downloaded from open-source sites like Hugging Face, is similarly largely visible and reviewable. (What isn’t available is the methods and data used to train the models.)
Large organizations should put any model core through their security testing and inspection processes, and they will also often post-train the models for their specific uses. So they work with, optimize, and understand the models before people start sending them prompts.
Could a model that is used for coding somehow throw malicious backdoors into the code it writes? Again, while that’s theoretically possible, it would require acrobatic feats to accomplish.
“There’s no reason that a sophisticated enough actor couldn’t train a model to be a completely amazing coding model in every circumstance, but when presented with a certain type of code base … some hidden training would kick in,” Atkins, who spends his days training models, postulated. But he adds: “I don’t know how you would do this.”
Because large language models are by nature creative, the odds are slim of getting a contemporary model to spit out malware in response to a preplanned perfect storm of context and prompt. Even slimmer are the chances that any enterprise would then use that code.
Could it happen in the future? That’s anyone’s guess. But enterprises are also building their AI apps to be model-agnostic and to use multiple models. So even if Chinese models are the best for the price today, enterprises won’t be locked into using them forever.
“I think instead of the conversation being about how to ban Chinese models, it should be about how do we foster a good, open ecosystem here in the U.S.,” Atkins says.
Arcee also gains advantages from Chinese models. Because they are open, the startup “benefits from those models being good because we can learn what they did. We can build on top of them. Then they can learn what we do,” he says. “We have tremendous respect for the people building those models, the individual researchers.”
Ultimately, the way to compete with Chinese models “is to release a model that is better,” says Atkins. “We need to give them something to talk about.”

### [E1398] TechCrunch AI | 2026-07-22 16:24 UTC | 盘中
Arcee, a US open source AI lab, says Chinese models are not inherently dangerous
URL: https://techcrunch.com/2026/07/22/arcee-a-us-open-source-ai-lab-says-chinese-models-are-not-inherently-dangerous/
As Chinese open-weight AI models grow in capability and popularity, arguments about what should be done about them have once again reached a fever pitch.
There’s talk that the Trump administration might try to ban them (though it hasn’t yet acted on the idea). Meanwhile, proprietary model makers, particularly OpenAI and Anthropic, appear increasingly concerned about them.
Open-weight models such as Moonshot AI’s Kimi K3 or Alibaba’s Qwen offer inference at a fraction of the token cost of closed-source models from these large U.S. labs. The fear is that they also pose some sort of threat. Certainly, they threaten the profit margins of the large proprietary AI labs.
But should enterprises running these models in their own data centers succumb to the fear that they could be a vector for Chinese hackers?
No, says Lucas Atkins, the CTO of Arcee, which is building open models to give U.S. companies a homegrown alternative to Chinese models.
If any startup would benefit from a ban on Chinese models, Arcee would. But Atkins says China’s open models are no more dangerous than any other open-source software a company may use. In fact, he says, they even offer benefits even to his own company.
“A lot of people view this as similar to a Chinese software program. Like, it was coded with these x, y, z intentions” that a bad actor could simply command, he said.
“That is fundamentally not how these models are trained. There is really not any way for an Arcee, or an Alibaba, to make a model, have someone run it in their own environment and for us have any access to it whatsoever,” he explained.
While most of these models are what’s known as “open weight,” and are not really fully open-source software, the source code (the part that will actually run on servers), if it is downloaded from open-source sites like Hugging Face, is similarly largely visible and reviewable. (What isn’t available is the methods and data used to train the models.)
Large organizations should put any model core through their security testing and inspection processes, and they will also often post-train the models for their specific uses. So they work with, optimize, and understand the models before people start sending them prompts.
Could a model that is used for coding somehow throw malicious backdoors into the code it writes? Again, while that’s theoretically possible, it would require acrobatic feats to accomplish.
“There’s no reason that a sophisticated enough actor couldn’t train a model to be a completely amazing coding model in every circumstance, but when presented with a certain type of code base … some hidden training would kick in,” Atkins, who spends his days training models, postulated. But he adds: “I don’t know how you would do this.”
Because large language models are by nature creative, the odds are slim of getting a contemporary model to spit out malware in response to a preplanned perfect storm of context and prompt. Even slimmer are the chances that any enterprise would then use that code.
Could it happen in the future? That’s anyone’s guess. But enterprises are also building their AI apps to be model-agnostic and to use multiple models. So even if Chinese models are the best for the price today, enterprises won’t be locked into using them forever.
“I think instead of the conversation being about how to ban Chinese models, it should be about how do we foster a good, open ecosystem here in the U.S.,” Atkins says.
Arcee also gains advantages from Chinese models. Because they are open, the startup “benefits from those models being good because we can learn what they did. We can build on top of them. Then they can learn what we do,” he says. “We have tremendous respect for the people building those models, the individual researchers.”
Ultimately, the way to compete with Chinese models “is to release a model that is better,” says Atkins. “We need to give them something to talk about.”

### [E2093] First Squawk | 2026-07-22 19:37 UTC | 盘中
NVIDIA CEO JENSEN HUANG: U.S. HAS NOTHING TO FEAR FROM CHINA'S OPEN-SOURCE AI MODELS; THERE IS NO SCENARIO WHERE CHINA RUNS U.S. COMPANIES OFF THE ROAD - AXIOS
URL: https://x.com/FirstSquawk/status/2080014374232850579
NVIDIA CEO JENSEN HUANG: U.S. HAS NOTHING TO FEAR FROM CHINA'S OPEN-SOURCE AI MODELS; THERE IS NO SCENARIO WHERE CHINA RUNS U.S. COMPANIES OFF THE ROAD - AXIOS

## 研究交付要求

这是一条已经从整场材料语义拆出的独立催化，不要重新把它压回大盘总评。核实美股异动原因和传播时序；用结构化供应链、内部 KB 与公司披露坐实每个亚洲候选；逐腿查真实行情和本地市场 price-in；候选不成立也要逐条给出有证据的 AVOID。