国际新闻 · 中文译文与原文对照

星期五 最新 1,000 篇 · 滚动浏览

科技

SpaceX的子公司Cursor告知立法者,正在考虑逐步淘汰中国的人工智能模型。

来源:CNBC Top News发布 上传时间 (多伦多时间)1 家来源 · 1 篇报道

中文译文

据CNBC报道,SpaceX的人工智能编程平台Cursor正在考虑因“地缘政治风险”而停止使用中国开发的AI模型。SpaceX于8月收购了Cursor,其股权价值约为600亿美元。Cursor的编程模型Composer 2是基于中国公司Moonshot AI的Kimi K2.5开发的。Cursor是美国采用中国AI模型的企业之一,这类模型今年的全球使用量激增,令华盛顿的立法者感到担忧。随着中国AI模型的不断改进,美中之间的AI霸权争夺在2026年达到了新的高度。不过,根据众议院国土安全委员会和众议院中国问题特别委员会在周五发布之前独家提供给CNBC的一份报告,Cursor现在计划“由于地缘政治风险,探索逐步淘汰由中国开发的开源模型”。这份报告是在对中国人工智能模型被纳入美国科技领域的情况进行数月调查后发布的,其中指出Cursor已将这些计划“告知相关委员会”。Airbnb和DoorDash也参与了关于其使用中国人工智能模型的调查。

SpaceX于6月宣布达成购买Cursor的正式协议,该交易于8月完成。SpaceX正在开发Grok系列模型,而Cursor曾参与这些模型的训练工作。

Cursor、SpaceX和DoorDash均未回应CNBC的置评请求。Airbnb拒绝置评。美国商务部工业与安全局也未立即回应置评请求。

中国人工智能模型的应用

报告发现,中国人工智能模型的成本、定制化程度以及接近前沿的性能水平推动了其应用,而美国目前缺乏在“市场所需的价格和性能水平上”具有竞争力的开源替代方案。报告称:“北京现在正试图将这些技术优势转化为全球市场份额以及对他国的长期依赖。”中国驻英国大使馆未对置评请求作出回应,中国外交部也未立即回复。

报告还提出了解决其所发现问题的建议。这些建议包括推出一项“国家级美国开放计算计划”,该计划整合公共计算能力、共享的评估基础设施,以及为训练目的而“合法使用”美国的先进模型。

报告还指出,10月发布的禁止DeepSeek进入联邦民用机构系统的命令应作为基础,进一步禁止在联邦信息系统中以及在执行联邦合同时使用“中国开发的AI模型”。此外,华盛顿方面还担心中国企业通过海外数据中心远程获取英伟达芯片,以及利用“蒸馏”技术取得优势——在这种技术中,新模型会借助更旧、更成熟的模型来辅助训练。美国商务部工业与安全局周一宣布了一项临时最终规则,旨在防止中国企业远程访问人工智能芯片。

原文

SpaceX's artificial-intelligence coding platform, Cursor, is considering moving away from using of Chinese AI models due to "geopolitical risk," CNBC has learned.

Cursor, which SpaceX acquired at an implied $60 billion equity value in August, had built its coding model Composer 2 on top of Chinese lab Moonshot AI's Kimi K2.5.

It's among the U.S. firms that have adopted Chinese models, whose usage has spiked across the globe this year, worrying lawmakers in Washington. The battle for AI supremacy between the U.S. and China has reached new heights in 2026 as Chinese models have improved.

But now Cursor plans to "explore phasing out PRC [People's Republic of China]-developed open-weight models because of geopolitical risk," according to a report produced by the House Committee on Homeland Security and the House Select Committee on China, shared exclusively with CNBC ahead of its publication on Friday.

The report, which follows a months-long investigation into the adoption of Chinese AI models into the American tech stack, said Cursor "informed the Committees" of those plans. Airbnb and DoorDash also engaged with the investigation about their use of Chinese AI models.

SpaceX announced it entered into a formal agreement to buy Cursor in June; the deal was completed in August. SpaceX is developing the Grok family of models, which Cursor has been involved in training.

Cursor, SpaceX and DoorDash did not respond to a CNBC request for comment. Airbnb declined to comment. The U.S. Commerce Department's Bureau of Industry and Security did not immediately respond to a request for comment.

Chinese AI adoption

The report found that cost, customization and near-frontier performance of Chinese AI models were driving adoption, with the U.S. lacking a competitive open-weight alternative at "price and performance the market demands."

"Beijing is now seeking to turn those technological gains into global market share and long-term dependence," the report says. China's embassy in the U.K. did not respond to a request for comment. China's Foreign Ministry did not immediately respond.

It also sets out recommendations to address the issues it has identified.

Those include launching a "national American open-weight initiative" that combines public computing power, shared evaluation infrastructure and "lawful access" to American frontier models for training.

The report also says an order, made public in October, that banned DeepSeek from federal civilian agency systems, should be built on, by prohibiting the use of "PRC-developed AI models" in federal information systems and in performance of federal contracts.

Separately, Washington is also concerned about Chinese companies accessing Nvidia chips remotely, via overseas data centers, and gaining ground by using "distillation," where new models use older, more established ones to help train them.

The U.S. Commerce Department's Bureau of Industry and Security, or BIS, announced on Monday an interim final rule designed to prevent Chinese companies' remote access to AI chips.