美國政府網站使用中國AI搜尋工具，FBI稱其複製Anthropic技術

作者：Courtney RozenBy Courtney Rozen

華盛頓，9月17日（路透社）——一個美國政府網站採用中國企業的AI模型，讓使用者搜尋擬議中的聯邦法規；然而美國聯邦調查局（FBI）先前指控該模型開發商阿里巴巴（Alibaba）「惡意」複製美國競爭對手Anthropic的技術。WASHINGTON, Sept 17 (Reuters) - A US government website employed an AI model from a Chinese company for users to search proposed federal regulations despite the FBI accusing the model's developer, Alibaba, of "malicious" copying of US rival Anthropic's technology.

由美國國家檔案暨紀錄管理局（National Archives）營運的《聯邦公報》（Federal Register）網站，週三提供使用阿里巴巴Qwen模型的搜尋工具， alongside 其他搜尋選項。該工具何時上線仍不清楚。The website of the Federal Register, which is run by the National Archives, had the search tool using Alibaba's Qwen model available on Wednesday along with other search options. It was unclear when it was deployed.

路透社檢視的截圖與該網站原始碼封存版本顯示，這項由Qwen驅動的搜尋功能週三遭下架，時間約與社群媒體出現相關貼文的時間相近。The Qwen-powered search was taken down on Wednesday around the same time social media posts appeared about it, according to screenshots and an archived version of the site's source code reviewed by Reuters.

FBI上週指控阿里巴巴複製Anthropic的模型，以打造自家人工智慧工具；在兩國領導人下週會面前，這項指控升高了華府與北京之間的緊張。The FBI last week accused Alibaba of copying Anthropic's models to build its own artificial intelligence tools, raising tensions between Washington and Beijing ahead of a meeting next week between the countries' leaders.

美國官員當時表示，中國AI企業正以「工業規模」從事對美國技術的「惡意」複製。US officials said then that Chinese AI companies were engaged in "malicious" copying of US technology on an "industrial scale."

FBI週四拒絕發表評論。美國國家檔案暨紀錄管理局與白宮未回應採訪請求。中國駐美大使館發言人表示，相關複製指控「毫無根據」。The FBI declined to comment on Thursday. The National Archives and the White House did not respond to requests for comment. A spokesperson for the Chinese embassy in Washington said the allegations of copying were "unfounded."

美國政府使用Qwen一事曝光之際，全球正重新審視AI安全，而美國國內也在辯論是否需要AI監管保障措施，或這類措施是否會使美國處於競爭劣勢。The US government's use of Qwen emerged amid a global reckoning about AI safety and a US debate over whether regulatory safeguards on AI are needed or would put the country at a competitive disadvantage.

AI專家表示，在《聯邦公報》網站使用這項中國工具，未必構成國家安全風險。但此事引發外界質疑：聯邦政府是否與自身對中國AI模型傳達的訊息相互矛盾。AI experts said the use of the Chinese tool on the Federal Register's website may not have posed national security risks. But it raised questions about whether the federal government was contradicting its own messages about Chinese AI models.

資訊科技與創新基金會（Information Technology and Innovation Foundation）主席丹尼爾・卡斯楚（Daniel Castro）表示，美中正激烈競爭以打造更優越的AI模型，而美國一個政府機關卻選擇使用中國製AI模型，兩者之間存在落差。Daniel Castro, president of the Information Technology and Innovation Foundation, said there was a disconnect between the fierce US-China competition to build superior AI models and a US agency choosing to use a Chinese-made AI model.

他說：「這是我見過最瘋狂的事情之一。」"It's one of the most insane things I've ever seen," he said.

《聯邦公報》網站每日提供最新擬議法規清單，並將Qwen模型作為瀏覽公眾意見的工具。The Federal Register website, which provides a daily list of newly proposed regulations, offered the Qwen model as a tool to browse public comments.

Qwen是開放權重模型，意指該系統的關鍵組成部分公開可用。開發者可以下載並修改模型，以用於特定用途，成本通常低於Anthropic、OpenAI等公司提供的封閉模型。Qwen is an open-weight model, meaning key components of the system are publicly available. Developers can download and modify the model for specific uses, often at lower cost than closed models from companies such as Anthropic and OpenAI.

美國眾議員、密西根州共和黨籍的眾院中國問題特別委員會主席穆勒納爾（John Moolenaar）說：「任何聯邦政府實體都不應使用中國AI模型。這麼做只會讓聯邦政府更加依賴中國AI模型，而這不符合國家利益。」“No federal government entity should use a Chinese AI model," said US Representative John Moolenaar, a Michigan Republican who chairs the House China Committee. "Doing so only makes the federal government more dependent on Chinese AI models and that is not in the national interest."

資安專家表示，開放權重模型可以在企業自有IT基礎架構上運行，讓組織對敏感資料擁有更大掌控權；這些資料否則可能會與外部供應商共享。Security experts say open-weight models can run on a company's own IT infrastructure, giving organizations greater control over sensitive data that might otherwise be shared with outside providers.

喬治城大學法學院教授阿努潘・錢德（Anupam Chander）表示，使用Qwen似乎未造成立即的網路安全風險，不過這取決於該模型的訓練方式。《聯邦公報》網站內容本來就是公開資訊，因此該模型並未處理敏感政府資訊。Georgetown Law Professor Anupam Chander said the use of Qwen did not appear to present an immediate cybersecurity risk, though that depends on how the model was trained. The Federal Register website's content is already public, so the model was not working with sensitive government information.

美國民主黨籍維吉尼亞州參議員、參議院情報委員會副主席馬克・華納（Mark Warner）在聲明中表示，風險取決於美國資料是否離開政府的安全邊界，並在阿里巴巴控制的系統上處理。Democratic US Senator Mark Warner of Virginia, vice chairman of the Senate Intelligence Committee, said in a statement that the risk depends on whether US data left the government's security boundary and was processed on Alibaba-controlled systems.

今年4月，美國眾院中國問題與國土安全相關委員會致函同樣使用Qwen的Airbnb，要求其提供使用細節；這是針對中國AI工具整合進美國人使用的平台所造成「國家安全風險」的調查之一。In April, the US House committees on China and homeland security sent a letter to Airbnb, which also uses Qwen, asking for details about its use as part of an investigation into the “national security risks created by the integration” of Chinese AI tools into platforms used by Americans.

（華盛頓Courtney Rozen報導；Alexandra Alper、Cynthia Osterman與Bill Berkrot編輯）(Reporting by Courtney Rozen in Washington, Editing by Alexandra Alper, Cynthia Osterman and Bill Berkrot)