更新时间：2026年3月5日 / 美国东部时间下午4:35 / CBS新闻
美国军方已正式将人工智能公司 Anthropic 列为供应链风险，一位五角大楼高级官员和一位直接了解情况的消息人士向 CBS 新闻证实，这一全面举措可能会切断其与军方合同的联系。
特朗普政府与 Anthropic——唯一一家部署在五角大楼机密网络上的 AI 公司——在 Anthropic 推动设立明确禁止美军使用其 Claude 模型对美国民众进行大规模监控或驱动完全自主武器的护栏问题上陷入僵局。五角大楼表示，它需要能够将 Claude 用于”所有合法目的”，并辩称这些 AI 使用方式已属不被允许。
国防部长彼得·黑格斯（Pete Hegseth）上周宣布 Anthropic 将被切断政府合同并列为供应链风险，但 Anthropic 直到周四才收到这一步骤的正式通知。
黑格斯称军方将在六个月内逐步淘汰 Anthropic。一位消息人士告诉 CBS 新闻，周四的风险认定中未提供停用 Claude 的时间表。
此前两位知情人士告诉 CBS 新闻，美军在上周末对伊朗的打击行动中已使用 Claude 模型，但其具体部署方式尚不清楚。
Anthropic 在首席执行官达里奥·阿莫代伊（Dario Amodei）告诉投资者他仍在与五角大楼”尝试缓和局势”两天后，获得了供应链风险认定。阿莫代伊在摩根士丹利会议上表示双方”共同点远多于分歧”，这一言论被 CBS 新闻独家获取的音频记录证实。
Anthropic 此前誓言将通过法律途径挑战任何将其列为供应链风险的企图，称此举”法律上站不住脚”，并警告这将为”任何与政府谈判的美国公司”树立”危险先例”。
阿莫代伊上周在接受 CBS 新闻采访时表示，他希望与军方合作保护美国国家安全利益，但公司在坚持护栏规定上立场坚定。他认为 AI 可能赋予政府”远超美国价值观的大规模监控权力”，且 AI 精确性不足，无法用于无需人类干预即可自主瞄准人员的武器。他指出，法律尚未跟上技术发展。
“我们有这两条红线，从第一天起就有。我们仍在为这些红线辩护，绝不会让步。”阿莫代伊强调。
五角大楼立场是，美军对美国民众进行大规模监控本身已属非法，而完全自主武器也已受国防部内部政策限制，因此无需在书面文件中额外限制 AI 的这些使用场景。
五角大楼首席技术官埃米尔·迈克尔（Emil Michael）上周在 CBS Newslate 采访中表示：”在某种程度上，你必须信任军方会做出正确选择。”但他同时指出”我们永远不会说无法通过书面方式约束公司”。
迈克尔透露，五角大楼上周提出了一项妥协方案，即书面确认限制大规模监控和自主武器的法律与政策。Anthropic 认为这些妥协”不充分”，称该提议”充斥着法律术语”，实则让军方可以无视护栏规定。
上周双方分歧进一步激化，特朗普政府官员指责 Anthropic 试图限制军方行动并将自身价值观强加给联邦政府。黑格斯称 Anthropic”伪善”，迈克尔称阿莫代伊有”神权情结”，特朗普则称该公司”激进左翼”和”觉醒派”。
特朗普政府设定上周五晚为最后期限，要求 Anthropic 同意让军方将 Claude 用于”所有合法目的”。双方仍存在巨大分歧，特朗普周五下令联邦机构立即停止使用 Claude，而国防部被给予六个月时间逐步淘汰该技术。
Anthropic 的竞争对手 OpenAI（以 ChatGPT 闻名）随后宣布已与军方达成协议。
一位五角大楼高级官员周四对 CBS 新闻表示：”从根本上说，这关乎一个原则：军方必须能够合法使用技术。任何供应商都不能通过限制关键能力的合法使用来干预指挥链，危及战士安全。”
阿莫代伊强烈批评特朗普政府的决定，称其”报复性且惩罚性”。
当被问及对特朗普的信息时，阿莫代伊表示：”我们所做的一切都是为了这个国家，为了支持美国国家安全。与政府分歧是美国最基本的权利，我们是爱国者，始终坚守美国价值观。”
Pentagon formally designates Anthropic a supply chain risk amid feud over AI guardrails
Updated on: March 5, 2026 / 4:35 PM EST / CBS News
The U.S. military has formally designated artificial intelligence firm Anthropic a supply chain risk, a senior Pentagon official and a source directly familiar with the situation told CBS News, a sweeping move that could cut it off from military contracts.
The Trump administration and Anthropic — the only AI company deployed on the Pentagon’s classified networks — are at an impasse over Anthropic’s push for guardrails that would explicitly ban the U.S. military from using its Claude model to conduct mass surveillance on Americans or power fully autonomous weapons. The Pentagon says it needs the ability to use Claude for “all lawful purposes,” and argues those uses of AI are already not allowed.
Defense Secretary Pete Hegseth announced last week that Anthropic would be cut off from its government contracts and designated a supply chain risk, but Anthropic had not received formal notification of that step until Thursday.
Hegseth said the military will phase out Anthropic over six months. A source told CBS News that no timeline for offboarding Claude was provided in Thursday’s designation.
The U.S. military has used Claude in its strikes on Iran that began last weekend, two sources familiar with the matter previously told CBS News. It’s not clear exactly how the artificial intelligence model is being deployed.
Anthropic received the supply chain risk designation two days after CEO Dario Amodei told investors he was still in talks with the Pentagon “to try to deescalate the situation. Amodei said at a Morgan Stanley conference that the two sides “have much more in common than we have differences,” according to audio exclusively obtained by CBS News.
Anthropic has previously vowed to legally challenge any attempt to label it a supply chain risk, calling the move “legally unsound” and warning it would set a “dangerous precedent for any American company that negotiates with the government.”
In an interview with CBS News last week, Amodei said he wants to work with the military to protect U.S. national security interests, but the company is standing firm in insisting on guardrails. He argued that AI could offer the government vast new surveillance powers that are “contrary to American values,” and AI isn’t precise enough to be used for fully autonomous weapons that target people without human input. In his view, the law hasn’t caught up with technology.
“We have these two red lines,” Amodei said. “We’ve had them from day one. We are still advocating for those red lines. We’re not going to move on those red lines.”
The Pentagon’s position is that it’s already illegal for the military to conduct mass surveillance on Americans, and fully autonomous weapons are already restricted by internal Defense Department policies, so there is no need to put restrictions on any of those uses of AI in writing.
Emil Michael, the Pentagon’s chief technology officer,said in an interview with CBS Newslate last week: “At some level, you have to trust your military to do the right thing.” But he also noted that “we’ll never say that we’re not going to be able to defend ourselves in writing to a company.”
Michael said last week the Pentagon offered a compromise that would acknowledge in writing the laws and policies that restrict mass surveillance and autonomous weapons. Anthropic called those compromises inadequate, saying the offer was “paired with legalese” that effectively let the military disregard the guardrails.
The disagreement grew increasingly bitter last week, with Trump administration officials accusing Anthropic of trying to restrict the military’s operations and impose its own values onto the federal government. Hegseth called Anthropic “sanctimonious,” Michael said Amodei has a “God-complex” and Mr. Trump called the company “radical left” and “woke.”
The Trump administration gave Anthropic a deadline of last Friday evening to agree to let the military use Claude for “all lawful purposes.” With the two sides still far apart, Mr. Trump on Friday ordered federal agencies to immediately stop using Claude, though the Defense Department was given up to six months to phase the technology out.
Anthropic rival OpenAI — known for ChatGPT — then announced that it had cut a deal with the military.
“From the very beginning, this has been about one fundamental principle: the military being able to use technology for all lawful purposes,” a senior Pentagon official told CBS News on Thursday. “The military will not allow a vendor to insert itself into the chain of command by restricting the lawful use of a critical capability and put our warfighters at risk.”
Amodei has strongly criticized the Trump administration’s decision, calling it “retaliatory and punitive.”
Asked by CBS News last week if he had a message for Mr. Trump, Amodei said “everything we have done has been for the sake of this country” and “for the sake of supporting U.S. national security.”
“Disagreeing with the government is the most American thing in the world,” he said. “And we are patriots. In everything we have done here, we have stood up for the values of this country.”
发表回复