Anthropic 在法庭上击败五角大楼

一名联邦法官阻止五角大楼将 Anthropic 列为国家安全供应链风险，裁定政府在这家 AI 公司拒绝放宽其 Claude 模型军方使用限制后，对其进行了非法报复。

据 Reuters 报道，美国地区法官 Rita Lin 认定，五角大楼的行为侵犯了 Anthropic 在第一修正案下的权利，且该公司未获得第五修正案所要求的程序性保护。

Lin 还认定，国防部长 Pete Hegseth 将 Anthropic 指定为供应链风险的决定违反了相关法律，且该决定武断、反复无常。她写道，政府不能以国家安全为由，惩罚一家在 AI 政策问题上持有特定立场的公司。

“对国家安全的空洞援引，并不是惩罚和报复政府批评者的空白支票。”Lin 写道。

这场争议始于 2 月，当时五角大楼要求 Anthropic 允许 Claude 用于一切合法的军事用途。Anthropic 拒绝取消对其技术用于针对美国人的大规模监控以及全自主武器的限制。该公司一直主张，当前 AI 系统尚不足以可靠地用于自主武器，而国内监控可能侵犯公民权利。

五角大楼随后将 Anthropic 列为供应链风险。这一分类旨在保护军用系统免受被认为可能遭外国对手渗透或破坏的公司影响。该指定还阻止 Anthropic 获得部分军事合同。The Associated Press 报道称，在 Anthropic 拒绝允许其技术被无限制用于军事用途后，政府指控该公司威胁国家安全。

美国司法部辩称，推动五角大楼作出决定的是 Anthropic 的合同限制，而非该公司对 AI 安全的观点。政府律师表示，这些限制可能导致军事行动中 Claude 的使用方式存在不确定性。

Lin 驳回了政府的立场，认定现有记录未能证明供应链风险指定具有有效依据。她还认定，在该指定生效前，Anthropic 未获得法律规定的提出异议机会。

该裁决撤销了五角大楼将 Anthropic 列入黑名单的决定，并阻止政府执行 Anthropic 所质疑的限制。据 Associated Press 报道，政府预计将提起上诉。

这起案件并非 Anthropic 与联邦政府之间唯一的法律纠纷。据 Reuters 报道，该公司还在华盛顿特区面临另一项尚待审理的诉讼，涉及五角大楼作出的另一项供应链风险指定，该指定可能影响其参与民用政府合同的资格。

Anthropic 对该裁决表示欢迎，并称仍愿意与政府合作推进 AI 在国家安全领域的应用。

“我们仍然专注于与政府开展富有成效的合作，利用 AI 服务我们的国家安全，让所有美国人都能从这项技术中受益。”该公司在一份声明中表示。

A federal judge has prevented the Pentagon from blacklisting Anthropic as a national-security supply-chain risk, ruling that the government unlawfully retaliated against the AI company after it refused to loosen restrictions on how its Claude models could be used by the military.

U.S. District Judge Rita Lin determined that the Pentagon's actions violated Anthropic's First Amendment rights and that the company was denied the procedural protections required under the Fifth Amendment, Reuters reported.

Lin also found that Defense Secretary Pete Hegseth's decision to designate Anthropic a supply-chain risk violated the governing law and was arbitrary and capricious. She wrote that the government's invocation of national security could not be used as a punishment against a company for its position on AI policy.

"The empty invocation of national security is not a blank check to punish and retaliate against government critics," Lin wrote.

The dispute began in February after the Pentagon demanded that Anthropic permit Claude to be used for all lawful military purposes. Anthropic refused to remove restrictions on using its technology for mass surveillance of Americans and fully autonomous weapons. The company has argued that current AI systems are not reliable enough for autonomous weapons and that domestic surveillance could violate civil rights.

The Pentagon responded by designating Anthropic a supply-chain risk, a classification intended to protect military systems from companies considered vulnerable to infiltration or sabotage by foreign adversaries. The designation also blocked Anthropic from certain military contracts. The Associated Press reported that the government accused Anthropic of threatening national security after the company declined to allow unrestricted military use of its technology.

The Justice Department argued that Anthropic's contractual restrictions, rather than the company's views on AI safety, drove the Pentagon's decision. Government lawyers said the restrictions could create uncertainty over how Claude could be used during military operations.

Lin rejected the government's position, finding that the record did not establish a valid basis for the supply-chain designation. She also found that Anthropic had not been given the required opportunity to challenge the designation before it took effect.

The ruling vacates the Pentagon's blacklisting decision and blocks the government from enforcing the restrictions challenged by Anthropic. The government is expected to appeal, according to Associated Press.

The case is not the only legal fight between Anthropic and the federal government. The company has a separate lawsuit pending in Washington, D.C., involving another Pentagon supply-chain designation that could affect its eligibility for civilian government contracts, Reuters reported.

Anthropic welcomed the ruling and said it remained willing to work with the government on national-security applications of AI.

"We remain focused on working productively with the government to harness AI for our national security so all Americans benefit from this technology," the company said in a statement.