Secretary of Defense Pete Hegseth attends a Cabinet meeting at the White House on April 10, 2025, in Washington, DC.

...

Newsweek Editors

By Newsweek Editors

0

Share

Newsweek is a Trust Project member

See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.


In April 2018, thousands of Google employees circulated a blunt internal letter about a Pentagon AI pilot called Project Maven. “We believe that Google should not be in the business of war,” it read. It said that machine-learning work meant to “interpret video imagery” could “improve the targeting of drone strikes” and urged the company to swear off “warfare technology.”

Eight years later, Washington is no longer asking Silicon Valley whether it wants to be in the business of "warfare technology." It is asking how quickly it can stop acting like it has a choice. On Tuesday, Reuters reported that Anthropic, maker of the Claude AI model, has “no intention of easing its usage restrictions for military purposes,” even after a Pentagon meeting in which Defense Secretary Pete Hegseth delivered an ultimatum: accept the government’s “all lawful purposes” terms by Friday at 5 p.m., or face “drastic action.” This action could include being labeled a supply-chain risk or having the Defense Production Act invoked to force changes to Anthropic’s rules.

This is framed as a debate about “woke” corporate ethics versus national security. China, meanwhile, has sidestepped that debate entirely. It doesn’t need to bully its tech companies in public. It has already built a vast procurement pipeline that draws tech companies into the military supply chain.

Read More on News

Washington is debating the guardrails. Beijing already has a timetable.

Common Knowledge

To Anthropic and its supporters in the tech industry, the Pentagon’s ultimatum is a state-sponsored jailbreak. In January, CEO Dario Amodei wrote that AI should support national defense “in all ways except those which would make us more like our autocratic adversaries.” And in Tuesday’s meeting, Anthropic told reporters it was continuing “good-faith conversations” about usage policy “to ensure Anthropic can continue to support the government’s national security mission in line with what our models can reliably and responsibly do.”

Others worry less about killer robots than bureaucratic creep. Axios noted that “existing mass surveillance law doesn’t contemplate AI,” and said that even if the Pentagon is already permitted to collect huge troves of information, AI could “supercharge” that authority in ways that make targeting civilians easier. The Washington Post reported that experts questioned whether the Defense Production Act could even be used the way Hegseth is threatening.

However, the Pentagon’s Undersecretary for Research and Engineering, Emil Michael, told DefenseScoop that the department “won’t permit companies supplying the technology to determine its rules for operation.” He was equally blunt: “You can’t have an AI company sell AI to the Department of War and [then] don’t let it do Department of War things,” he said, adding that guardrails must be “tuned for military applications…so long as they’re lawful.”

The dispute also has a distinctly Trumpian edge. The Financial Times reported administration figures criticizing the company as “woke.”

Uncommon Knowledge

The Pentagon–Anthropic fight is being sold as an issue of morality. But alongside morality, there is capacity.

Until recently, Anthropic was the only major AI provider on American classified networks. Now the Pentagon has also reached an agreement with xAI, maker of Grok, to deploy across classified networks as well, as part of a broader push to bring multiple AI firms into classified environments. Even if Anthropic wins this battle on ethics, it may lose the war, because its leverage comes from being uniquely embedded.

China, by contrast, is building global leverage by making AI procurement normal. A report from September 2025 by CSET, the Center for Security and Emerging Technology at Georgetown University, illustrates the scale of this effort.

Researchers analyzed 2,857 AI-related PLA award notices published between January 2023 and December 2024. They found 2,090 of those notices included contract values totaling $535.5 million. That total is “a small fraction” of China’s reported military spending, CSET said. But the point isn’t the dollar figure. The point is the habit: thousands of posted procurements, spread across a wide range of suppliers, for capabilities CSET defines broadly to include “language and vision models,” “autonomous vehicles,” “test and evaluation technologies,” “data collection and processing systems,” “smart manufacturing and robotics,” and more.

The drive is accelerating. From May to December 2023, CSET counted 1,039 AI-related award notices. From May to December 2024, it counted 1,249, a 16 percent increase, and the value of those notices rose from $162.8 million to $183.2 million.

And the supplier base is not just state-owned behemoths. CSET says Beijing aims to “foster stronger ties between the civilian and defense sectors” and reduce barriers so civilian entities can compete for defense contracts. In the AI subset CSET analyzed, “nontraditional vendors” mattered: they won 764 contracts. That figure alone is the mirror image of Amodei’s “autocratic adversaries” warning.

The Pentagon may be bullying Anthropic. But China doesn’t need to bully in public because it has something Washington increasingly lacks: administrative confidence. America can't win an arms race by arguing about terms of service.


If you’re enjoying Uncommon Knowledge, please share. If you have suggestions for future editions or feedback, email subscriber.feedback@newsweek.com. We want to hear your voice.

View 0 Comments

Request Reprint & Licensing

Submit Correction

View Editorial & AI Guidelines