Anthropic Joins A.I. Giants to Provide Models to US Defense Agencies

Anthropic, a competitor to OpenAI that has positioned itself as a more safety-conscious alternative, announced today (Nov. 7) that it will provide U.S. defense and intelligence agencies with access to its Claude A.I. models. The deal will see Anthropic team up with Amazon Web Services (AWS) and Palantir to incorporate Claude in government efforts like data processing and document preparation.

Anthropic joins a growing trove of A.I. companies supplying their technologies to the U.S. government. Between August of 2022 and 2023, the value of A.I.-related federal contracts skyrocketed 150 percent to $675 million, according to a March report from the Washington, D.C.-based think tank Brookings Institute. The U.S. Department of Defense (DoD) is emerging as one of the most dominant players in this new space, seeing the value of its A.I.-related contracts jump from $190 million to $557 million during this time.

Under the deal, Anthropic’s Claude family of models will be available to government customers through Palantir’s platform on AWS. This access “will equip U.S. defense and intelligence organizations with powerful A.I. tools that can rapidly process and analyze vast amounts of complex data,” said Kate Earle Jensen, head of sales and partnerships at Anthropic, in a statement. “This will dramatically improve intelligence analysis and enable officials in their decision-making processes, streamline resource intensive tasks and boost operational efficiency across departments,” she added.

According to Anthropic’s usage policy, the company is allowed to enter into contracts with “carefully selected government entities” pursuing “foreign intelligence analysis in accordance with applicable law.” Use of its A.I. systems for disinformation campaigns, weapon design, censorship, domestic surveillance and “malicious” cyber operations, however, remain restricted.

Big Tech strengthens ties with DoD on A.I. application

Anthropic’s deal follows a similar announcement from Meta (META), which earlier this week unveiled plans to allow U.S. government agencies and contractors like Palantir, Lockheed Martin and Booz Allen to use its open-source Llama A.I. models for defense and national security applications. Meta’s models, which will be used to synthesize documents, accelerate code generation and strengthen cyber defense, will also be made available to agencies across the Five Eyes intelligence alliance, which consists of the U.S., Canada, Britain, Australia and New Zealand.

Microsoft, meanwhile, teamed up with Palantir in August to offer A.I. software and capabilities to defense-focused U.S. federal agencies. OpenAI, which earlier this year inked a deal with the government contractor Carahsoft, is also reportedly interesting in working with the DoD and the Department of Homeland Security.

These growing ties between Big Tech’s A.I. products and military applications have not been without scrutiny.

In 2018 thousands of Google employees protested against the company’s efforts in a Pentagon program known as Project Maven that used A.I. to improve analysis of drone strike videos, leading Google to end its contract. However, the tech company remains engaged in various contracts with defense agencies—something that became a point of contention earlier this year, when nearly 200 Google DeepMind employees reportedly signed a letter against such partnerships over fears that the A.I. lab’s technologies were being used for manufacturing weapons.