Pentagon Adds Nvidia, Microsoft, Amazon, and Reflection AI to Classified AI Push

The Pentagon has signed agreements with Nvidia, Microsoft, Reflection AI, and Amazon to use their AI tools and infrastructure on classified programs, bringing the total number of AI providers with classified status to seven.

OpenAI, SpaceX through xAI, and Google were already part of this agreement, which allows the Department of Defense to deploy their technology for “any lawful use.” It will gain access to AI models from SpaceX, OpenAI, Google, and Nvidia, while also using AI infrastructure from Amazon Web Services and Microsoft Azure.

It may also be able to access private alpha or beta builds of Reflection AI models, as the company has not yet released a model publicly.

Anthropic fallout expands Pentagon’s AI roster

The Pentagon has greatly expanded the number of companies with classified status following disagreements with Anthropic.

This led the US government to label the company a supply chain risk. The term “any lawful use” was the crux of the disagreement, as Anthropic CEO Dario Amodei said it did not go far enough and would not prevent the Department of Defense from using its AI model, Claude, for mass domestic surveillance or lethal autonomous weapons.

This led the Pentagon to cancel a $200 million contract with Anthropic and label it a supply chain risk, the first time this designation has been applied to a US-based company. At the time, Anthropic was the only AI model developer with classified status, meaning the Pentagon had to quickly sign agreements with rivals such as OpenAI.

Anthropic sued the US government in response, claiming it had lost millions in contracts through the designation.

Even with these disagreements, it is unclear whether the US government stopped using Anthropic. It was reportedly a key AI model in some of the planning for the Iran War, and Anthropic has briefed several members of the Trump Administration on its newest model, Mythos. 

The National Security Agency, which sits within the Department of Defense, has reportedly gained access to Mythos and is testing it against its own infrastructure and Microsoft systems.

Anthropic is not the only one unhappy

While Anthropic might have the only CEO publicly unhappy with the Pentagon, plenty of employees at these companies are making their voices heard. Around 600 employees in Google’s AI and Cloud divisions signed a letter to CEO Sundar Pichai urging him to end the deal.

Googlers have been outspoken about government contracts in the past. In 2018, a 4,000-strong petition to end a drone analytics program called Project Maven led to the contract not being renewed. In the same year, Google withdrew from bidding on a cloud project tied to the Pentagon. 

In more recent years, Google leadership has fought back against employee opposition, firing around 50 employees who protested a cloud project between Google and the Israeli government.

OpenAI has had its own internal disagreements, with a senior hardware leader at the company quitting after it signed an agreement with the Pentagon. A similar employee letter to CEO Sam Altman was sent by hundreds of employees, but no action has yet been taken to change or terminate their contract with the Pentagon.

Also read: NSA testing of Anthropic’s Mythos model has raised questions about government AI access, cybersecurity testing, and the limits of military use.

The post Pentagon Adds Nvidia, Microsoft, Amazon, and Reflection AI to Classified AI Push appeared first on eWEEK.