5 Big Questions from the Humanity vs. Department of Defense debate: ‘It’s all very puzzling’

Defense Secretary Pete Hegseth’s decision Friday to classify Anthropic as a “supply chain risk to national security” raises more questions than answers.

“It’s all very puzzling,” Herbert Linn, a senior fellow at Stanford University’s Center for International Security Cooperation, said in an interview with CNBC.

Anthropic is the only U.S. company to be publicly designated as a supply chain risk, as supply chain risk designations have traditionally been used against foreign adversaries. However, the company has not received any formal declaration beyond the social media post.

Formal designation requires defense vendors and contractors to certify that they are not using Anthropic models in their work with the Department of Defense.

The controversy centered on how Anthropic’s artificial intelligence model might be used in the military. The Pentagon wanted Anthropic to have unfettered access to its Claude models for all lawful purposes, but Anthropic wanted assurances that its technology would not be used for fully autonomous weapons or domestic mass surveillance.

After failing to reach a deal by Friday’s deadline, President Donald Trump directed federal agencies to “immediately cease” all use of Anthropic’s technology and said agencies like the Department of Defense would have a six-month phase-out period.

Experts told CNBC that the supply chain risk designation is highly unusual, especially since the US and Israel launched an attack on Iran just hours later. A group of retired defense officials, policy leaders and executives sent a letter to Congress on Thursday defending Antropic and calling the Trump administration’s designation a “dangerous precedent.”

As CNBC previously reported, Anthropic’s models are still being used to support U.S. military operations in Iran even after the company was blacklisted.

Talks between Anthropic and the Pentagon have reportedly resumed, according to the Financial Times, but as of Thursday, big questions remained on the issue.

Why is the US government still using Claude?

Stanford University’s Lin doesn’t understand why the Department of Defense is still using Anthropic’s models in classified environments if they pose such a threat. He said if the Trump administration truly views Antropics as a risk to national security, it makes no sense to phase out the model over the long term.

“OK, wait a minute, are we going to use them for another six months because they’re such a dangerous player to the national security of the United States? Huh?” Lin said.

Michael Horowitz, a senior fellow for innovation at the Council on Foreign Relations, said it was “particularly noteworthy” that Anthropic’s model was used to support U.S. military operations in Iran. He said there was “no clearer indication” of how much the Pentagon values ​​the technology.

“Even in the face of such a bitter feud between the company and the Department of Defense, they still use their technology for some of the most important military operations the United States is conducting,” he said.

The transition from Anthropic to a new vendor is time consuming and comes at a significant cost in terms of efficiency, said Jacqueline Schneider, a Hargrove Hoover fellow at Stanford University’s Hoover Institution.

Until recently, Anthropic was the only AI company approved to deploy models across sensitive government networks. OpenAI and Elon Musk’s xAI have been certified, but their systems cannot be deployed or adopted overnight.

What is the real threat?

Lin said that by designating Anthropic as a supply chain risk, the Department of Defense is indicating that the company is “very bad” for U.S. national security. But he stressed that authorities have not clearly explained what kind of threat the company poses.

“They haven’t pointed out any technical flaws, they haven’t pointed out any hacking,” Lin said. “They say things like, ‘They’re arrogant,’ or ‘We don’t want the Department of Defense to tell us what to do in a hypothetical situation that hasn’t happened yet.'”

Lin said the other punishment Hegseth is threatening to impose on Anthropic by invoking the Defense Production Act is also inconsistent with the idea that the company poses a threat to national security.

The Defense Production Act allows the president to use emergency powers to control domestic industry in the interest of national security. That could essentially force Anthropic to use its technology for the Pentagon.

Horowitz said he believes the clash between Anthropic and the Pentagon is a “sham” for a policy debate.

Months earlier, venture capitalist and White House kingpin of AI and cryptocurrencies, David Sachs, criticized the company for “pursuing a sophisticated regulatory acquisition strategy based on fear-mongering” following an essay published by executives, and conservatives have repeatedly accused Anthropic of promoting “woke AI.”

Anthropic CEO Dario Amodei took a different approach from other tech executives, avoiding early cozying up to the Trump administration.

“This feels to me like a debate about politics and the personal,” Horowitz said.

Is an official designation coming soon?

Anthropic is not designated as a supply chain risk by any official measure, and there are open questions about whether and when the company should expect risk. Defense contractors will have to decide whether to follow Hegseth’s instructions on social media or wait for more formal guidance.

Multiple executives told CNBC that their companies are moving away from Anthropic’s model, and one venture capitalist said many portfolio companies are making the switch “out of an abundance of caution.” But other stakeholders, including C3 AI chairman Tom Siebel, said they saw no “need to mitigate” the technology “until litigation occurs.”

Schneider said companies are rational and if they think it is risky to partner with Anthropic, whether or not it is formally declared a supply chain risk, they will hedge and look for other partners.

“There are all kinds of decisions that were made within the Trump administration that need to be more codified in law,” Schneider said. “Even the example of moving from the Department of Defense to the (Department of the Army). This needs to be further codified by law, but all contractors are using the Department of Defense.”

Still, Sameer Jain, vice president of policy at the Center for Democratic Technology, said a social media post alone likely isn’t enough to actually trigger a designation.

“There is a process that this law requires, including an actual investigation that if Anthropic is part of the supply chain, it poses a national security risk,” he said in an interview. “Actually, I don’t think that predicate is satisfied here.”

Anthropic said in a statement Friday that it “will challenge the supply chain risk designation in court.”

Does this have something to do with the US attack on Iran?

For Schneider, the fight between Anthropic and the Pentagon means war in Iran is on the horizon. She said she wonders if the two conflicts are happening in parallel or if they are somehow related.

“It’s clear that right before we go to war, we’re not going to walk away from technology that’s deeply embedded in the wartime process,” Schneider said.

She said she was surprised that the Pentagon was willing to expend such a “staggering amount of energy” on a public confrontation ahead of the first attack, since planning a military operation of that magnitude would have required “many sleepless nights.”

What happens next?

As the war in Iran enters its sixth day, Anthropic’s future with the Department of Defense remains a big mystery.

Horowitz said the six-month suspension would provide “some kind of reconsideration” within the Pentagon, especially since there is so much interest in the dispute from members of Congress and the broader public market.

Lin expressed similar sentiments, saying he wouldn’t bet on Anthropic’s model leaving the Pentagon after a year.

Schneider isn’t so sure.

“I wish we had thought more clearly about what was going to happen with this, but this is all unprecedented,” she said. As for historical or similar examples, Schneider said, “I don’t have any of those. It’s just very limited.”

The Pentagon declined to comment. Antropic has not commented.

WATCH: Anthropic’s annual revenue exceeds $19 billion