|
|
|
Mar 02, 2026
|
|
|
|
|
Supported by
|
|
|
|
|
|
|
Welcome back! The Pentagon declares Anthropic a supply chain risk, cutting it off from military contractors, and signs a deal with OpenAI. Amazon agrees to invest up to $50 billion in OpenAI.
|
|
|
|
Defense Secretary Pete Hegseth on Friday declared Anthropic a supply chain risk, an extraordinary sanction against an American company that is usually reserved for foreign adversaries. “Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic,” Hegseth wrote. The exact scope of Hegseth’s decision was not immediately clear. Hegseth did not specify whether those contractors would only be banned from using Anthropic in their work with the military. Instead,
Hegseth’s statement appeared to bar Anthropic from “any commercial activity” with any company that works with the military. It’s unclear, for example, whether Anthropic would no longer be able to work with its primary cloud providers, Google and Amazon, which both hold defense contracts. Anthropic, Google and Amazon did not immediately respond to requests for comment. Hegseth’s statement follows an announcement from President Donald Trump earlier Friday that directed all federal agencies to stop using Anthropic’s technology. Hegseth said that
Anthropic’s relationship with the federal government and the United States armed forces had been “permanently altered,” and that “this decision is final.”
|
|
|
|
OpenAI CEO Sam Altman said late Saturday that the company’s recent agreement to supply the Pentagon with AI models would fall within the company’s own “redlines” limiting how it’s used in military situations, but he argued that it was appropriate the government—not the company—would have the final say in some situations. “Seems fine for us to decide how ChatGPT should respond to a controversial question,” he wrote on X in response to a follower’s question. “But I really don’t want us to decide what to do if a nuke is coming towards the U.S.” The discussion followed the company’s announcement late Friday that it had reached an agreement with the Pentagon about using the company’s models for classified work. The announcement, which Altman acknowledged was “rushed,” followed a move by the Defense Department to sanction Anthropic after the AI company said it would not agree to have its AI be used for domestic mass surveillance and fully autonomous weapons systems. Altman said he expected the military to uphold the laws that prevent the activities, such as domestic mass surveillance, that OpenAI’s corporate policy restricts.
|
|
|
|
Amazon confirmed it would invest up to $50 billion in OpenAI, in two stages, as part of a broad agreement with the ChatGPT maker that will expand OpenAI’s existing $38 billion agreement to rent servers on Amazon’s cloud service AWS by $100 billion and will involve OpenAI using 2 gigawatts of Amazon’s Trainium AI chips through AWS. Amazon’s cloud service, AWS, also will be the “exclusive third party cloud distribution provider” for OpenAI’s new enterprise focused AI management service, OpenAI Frontier, the companies said. The details in the agreement were in line with what The Information has previously reported. That includes Amazon investing $15 billion by the end of March and another $35 billion “in the coming months when certain conditions are met.” A securities filing by Amazon said Amazon’s obligations to buy the extra shares was based on the earlier of OpenAI “meeting specified milestones” and OpenAI going public. The Information previously reported the specified milestones related to OpenAI achieving artificial general intelligence, generally defined as AI on a par with human abilities. But the formal contract, included in Amazon’s securities filing, doesn’t make clear what the milestones are.
|
|
|
|
Nvidia is expected to announce a new chip later this month using technology from startup Groq, The Wall Street Journal reported late Friday. The chip is aimed at speeding up inference computing, the term for running AI on servers after AI models have been fully developed, the newspaper said. The industry’s demand for AI chips is gradually shifting from training AI models to operating the models for users. The chip is set to be revealed at Nvidia’s GTC conference in San Jose in mid-March, the Journal reported. Nvidia last year agreed to pay $20 billion to license Groq’s technology and hire its leadership team. The new chip would be a break from its normal strategy, which is to release chips that are more versatile to cater to the widest possible market. The company typically touts chips that are good at both training and inference, though companies like OpenAI have reportedly been looking for faster alternatives, believing that dedicated inference chips would be better for their operations. OpenAI has agreed to be one of the biggest customers for the new chip, the Journal added.
|
|
|
|
President Donald Trump said on Friday afternoon that he was directing every U.S. federal agency to stop using Anthropic’s technology, further escalating a feud between Anthropic and the Pentagon over Anthropic’s refusal to drop its AI safeguards. Trump’s statement came an hour before a deadline for Anthropic to agree to the Pentagon’s demands to make its AI available for “any lawful use” or face consequences. Defense Secretary Pete Hegseth had previously threatened to declare Anthropic a supply chain risk, which would have cut Anthropic
off from the military supply chain, or use the Defense Production Act to force Anthropic to tailor its AI models for the Pentagon. Anthropic did not immediately respond to a request for comment. Trump said that there would be a six month phase out period for agencies using Anthropic’s policies. “Anthropic better get their act together, and be helpful during this phase out period, or I will use the Full Power of the Presidency to make them comply, with major civil and criminal consequences to follow,” Trump wrote. Trump’s statement raises questions about how his administration will deal with other AI companies, especially OpenAI. OpenAI CEO Sam Altman told employees on Thursday that OpenAI shares Anthropic’s same red lines—refusing to let their AI be used for mass domestic surveillance or fully autonomous lethal weapons—and was working to negotiate a
deal with the Pentagon under those terms. An OpenAI spokesperson did not immediately have a comment.
|
|
|
|
| |
