Federal appeals court rules Pentagon’s blacklist of Anthropic was legal

Carlos Barria/File/Reuters

The Department of Defense can blacklist Anthropic’s artificial intelligence technology, a federal appeals court ruled on Friday. The decision is a blow to Anthropic after it won a separate challenge in a California court in August that partially invalidated the Trump administration’s label.

The Trump administration earlier this year labeled Anthropic a “supply chain risk” and ordered federal agencies and contractors that work with the military to cease business with the company. The move came after Anthropic refused to allow the Pentagon to use its AI for “all lawful purposes” as the department wanted, citing the company’s red lines around autonomous weapons and surveillance of US citizens.

Friday’s decision from the DC Circuit means that Claude will still be prohibited within the Pentagon, although the earlier California ruling allows other government agencies and contractors to continue work with Anthropic. The government has not appealed the California ruling.

The designation from the Pentagon put a significant portion of Anthropic’s enterprise business in jeopardy , meaning that no contractor, supplier or partner that works with the military would be allowed to conduct business with Anthropic — including using its paid version of Claude. The company is betting on its strength in enterprise AI to fuel a potentially trillion-dollar initial public offering , which could take place before the end of this year.

Anthropic sued over the designation, accusing the government of violating the First Amendment and calling the supply chain risk designation — usually reserved for firms associated with foreign adversaries and never before used on an American company — legally unsound.

But the federal Court of Appeals for the District of Columbia rejected Anthropic’s arguments, ruling the Pentagon had “ample support” for its finding that integrating the company’s Claude AI into its systems could pose a national security threat. One of the three judges on the panel dissented, arguing that there was not a legitimate concern that Anthropic could manipulate its technology in order to pose a security risk.

An Anthropic spokesperson said the company disagrees with Friday’s decision.

“Another federal court has already held the government’s parallel designation unlawful,” the spokesperson said in a statement. “We remain confident in our position and are considering all options, including further review.”

Anthropic could choose to ask the entire DC appeals court rehear the case “en banc,” which would force a ruling from the 11 judges that make up the full court, rather than the three-judge panel that ruled Friday. It can also choose to go straight to the Supreme Court.

Because of the multiple designations the Pentagon used to issue the label, the DC and California courts’ rulings were based on different statutory authorities. The DC court did not reverse the California court ruling.

The Pentagon celebrated the ruling, with Undersecretary Emil Michael writing on X that “The hammer of justice has smashed @AnthropicAI arguments… Warfighters will sleep better knowing that no private company will insert their opinions in the chain of command.”

卡洛斯·巴里亚/资料图/路透社

联邦上诉法院周五裁定，国防部可以将人脑公司（Anthropic）的人工智能技术列入黑名单。此前，该公司在加州法院赢得了一场独立诉讼，部分推翻了特朗普政府的禁令，此次裁决对人脑公司而言无疑是一次打击。

今年早些时候，特朗普政府将Anthropic公司列为“供应链风险”，并命令与军方合作的联邦机构和承包商停止与该公司的业务往来。此前，Anthropic公司拒绝五角大楼将其人工智能技术用于“所有合法用途”，理由是该公司在自主武器和监视美国公民方面划定了红线。

周五哥伦比亚特区巡回上诉法院的裁决意味着克劳德仍将被禁止在五角大楼内工作，尽管此前加州的裁决允许其他政府机构和承包商继续与安人智公司合作。美国政府尚未就加州的裁决提出上诉。

五角大楼的这项认定使Anthropic公司的大部分企业业务岌岌可危，这意味着任何与军方合作的承包商、供应商或合作伙伴都将被禁止与Anthropic公司开展业务，包括使用其付费版的Claude人工智能系统。该公司正押注其在企业人工智能领域的实力，以推动其规模可能达到万亿美元的首次公开募股（IPO），IPO有望在今年年底前完成。

人类公司就此认定提起诉讼，指责政府违反了宪法第一修正案，并称供应链风险认定（通常保留给与外国敌对势力有关联的公司，以前从未用于美国公司）在法律上站不住脚。

但哥伦比亚特区联邦上诉法院驳回了Anthropic公司的论点，裁定五角大楼有“充分证据”支持其关于将该公司Claude人工智能系统整合到其系统中可能构成国家安全威胁的结论。合议庭三名法官中的一位持不同意见，认为Anthropic公司可能操纵其技术以构成安全风险的担忧并不合理。

人类组织发言人表示，该公司不同意周五的决定。

发言人在一份声明中表示：“另一家联邦法院已经裁定政府的平行指定是非法的。我们仍然对我们的立场充满信心，并正在考虑所有选项，包括进一步审查。”

Anthropic可以选择要求哥伦比亚特区上诉法院全体法官进行“全席审理”（en banc），这将迫使由11名法官组成的全体法官小组做出裁决，而不是由周五做出裁决的三名法官组成的小组。它也可以选择直接向最高法院上诉。

由于五角大楼在发布标签时使用了多种不同的名称，华盛顿特区法院和加利福尼亚州法院的裁决依据的法律依据各不相同。华盛顿特区法院并未推翻加利福尼亚州法院的裁决。

五角大楼对这一裁决表示庆祝，副部长埃米尔·迈克尔在X上写道：“正义之锤粉碎了@AnthropicAI的论点……得知没有私营公司能够将他们的意见强加于指挥系统，作战人员可以睡得更安稳了。”