法院支持五角大楼将Anthropic列为供应链风险的决策

美国上诉法院裁定支持五角大楼将Claude排除在系统之外，Anthropic面临IPO前的新挫折

近日，美国一家联邦上诉法院正式确认了五角大楼将人工智能模型Claude排除在其部分系统和合同之外的决定。这一新的法律挫折削弱了Anthropic在首次公开募股（IPO）前的势头——该IPO预计将使公司估值达到2万亿美元。然而，这并不意味着法律战已经结束。

核心要点

• 美国联邦上诉法院确认了Claude被排除在特定五角大楼系统和合同之外的合法性。
• 冲突的核心在于Anthropic坚持的两项限制：禁止开发自主武器和禁止大规模监控。
• 目前存在两份相互对立的司法裁决，Anthropic仍有多种上诉途径可供选择。
• 此次挫折发生在公司筹备潜在IPO之际，市场传闻其估值约为2万亿美元。
• Anthropic以2:1的票数劣势败诉，但仍在研究后续法律行动。

2:1多数意见维持原判

9月25日，哥伦比亚特区联邦上诉法院驳回了Anthropic提出的上诉申请，尽管该公司此前曾获得过暂缓执行的许可。三名法官中有两名认为，五角大楼有权将Anthropic列为“供应链风险”实体。

The ruling allows the Department of Defense to continue removing Claude from the system and prohibits military subcontractors from using Anthropic products while working for the Pentagon.

judgment highlight the importance of the defeat:

• The decision was passed by a majority of two judges, with one judge dissenting;
• Claude must be removed from the relevant military system;
• The restrictions are mainly for contracts related to the Pentagon;
• Anthropic is still evaluating possible legal next steps.

Judge Gregory Katsas argued that the Department of Defense had "good reason" to conclude that continued integration of Claude could pose a risk covered by the law. The court also rejected Anthropic's defense based on the rights to freedom of speech and due process.

Two restrictions spark conflict

Anthropic has previously allowed the U.S. government to use Claude for foreign intelligence analysis, military system design, and the execution of certain cyber offensive operations. However, the company insists on retaining two bans:

1. Fully autonomous lethal weapons: Ban the development of weapon systems that can select and attack targets without human intervention.
2. Mass surveillance: Mass surveillance of U.S. citizens is prohibited.

At the same time, the Pentagon is demanding a contract that allows all legal uses. Negotiations broke down in February last year. CEO Dario Amodi once said through on-chain data that autonomous weapons are not reliable enough, and large-scale surveillance, even if legal, is not in line with democratic values.

The Department of Defense is mainly concerned that Anthropic may change Claude's security measures, thereby interfering with ongoing operations. In fact, the court noted that the model had previously rejected a number of government requests, including requests involving confidential documents and sensitive health research.

As a result, the Pentagon invoked a 2018 federal law aimed at ensuring supply chain security. The legislation allows agencies to exclude service providers when they believe they are monitoring, disrupting or modifying their operations.

Two opposing judicial decisions coexist

The September 25 decision does not automatically revoke Anthropic's victory in the August 27 California lawsuit. In fact, federal judge Rita Lin previously ruled that another Pentagon measure was unlawful retaliation and made clear that the company did not receive the legal protections it needed.

The court in Washington reviewed an independent process that relied largely on the 2018 law. The majority opinion is that the Pentagon has properly notified Anthropic and given it the opportunity to refute his exclusion.

Judge Karen Henderson objected. She believes that the majority opinion gives the government too much power to force technology providers to accept their terms under the pretext of "national threats."

Anthropic reiterated that California's verdict remained in its favor. The company said it was evaluating its next options and might request a retrial of the case or a lawsuit to the Supreme Court, but has not yet announced a specific strategy.

The US$2 trillion IPO remains hypothetical

This setback comes at a critical time as Anthropic prepares to enter the stock market. The company formally filed a confidential draft Form S-1 with the Securities and Exchange Commission (SEC) on June 1. However, the company has not yet confirmed the issue price, target amount or future valuation.

The IPO financing could be as high as US$100 billion, valuing Claude's creator Anthropic at about US$2 trillion. Nvidia is rumoured to be discussing an investment that could reach $10 billion, but the two sides have not confirmed those talks.

Anthropic is valued at US$965 billion after completing a new round of financing in May. If it reaches US$2 trillion, it means that its market value will more than double in just a few months.

While exclusion from the military market does not directly threaten all of its commercial activities, it increases regulatory risks that future investors will need to assess, especially in terms of public contracts and relationships with defense suppliers.