Anthropic and Pentagon Resume Talks After Week-Long Standoff

Anthropic CEO Dario Amodei has resumed negotiations with the Pentagon over the terms governing military access to Claude, one week after the Trump administration ordered every federal agency to stop using the company's technology and Defense Secretary Pete Hegseth branded Anthropic a supply chain risk to national security.

The Financial Times reported Wednesday that Amodei is in direct talks with Emil Michael, the under-secretary of defense for research and engineering, in what multiple outlets are calling a last-ditch effort to avoid permanent blacklisting.

TL;DR

Anthropic CEO Dario Amodei confirmed at a Morgan Stanley conference he's trying to "deescalate" the Pentagon standoff
Amodei is negotiating directly with under-secretary Emil Michael, the former Uber CBO who called him a "liar" last week
The core sticking point: a clause about "analysis of bulk acquired data" that Anthropic says would enable mass domestic surveillance
The Pentagon offered to accept Anthropic's terms if the company deleted that one phrase - Amodei refused
Lawfare legal scholars say the supply chain risk designation "won't survive first contact with the legal system"
The White House has separately cast doubt on whether reconciliation is possible

What Changed

The Morgan Stanley Appearance

CBS News obtained audio of Amodei's remarks at the Morgan Stanley Technology, Media and Telecom Conference in San Francisco on Tuesday. Speaking to an investor audience, Amodei struck a conciliatory tone - a marked shift from the internal memo last week where he accused OpenAI of "straight up lies" about its own Pentagon deal.

"We have much more in common than we have differences."

Amodei said Anthropic is trying to "deescalate the situation and come to some agreement that works for us and works for them." He framed the company's defiance as patriotic: "Disagreeing with the government is the most American thing in the world."

The company's position has not changed on substance. Anthropic still refuses to allow Claude to be used for mass surveillance of Americans or for fully autonomous weapons. But the language has shifted from confrontation to negotiation.

The Negotiating Partner

The man across the table is Emil Michael, a former Uber chief business officer who joined the Pentagon earlier this year. Michael has been blunt about his position. He told reporters last week: "You can't have an AI company sell AI to the Department of War and not let it do Department of War things."

He also publicly called Amodei a "liar" and accused him of having a "God complex" - unusual language between a government official and a CEO of a company he's simultaneously trying to bring back to the table.

Michael's background matters. At Uber he was Travis Kalanick's top lieutenant during the company's most aggressive expansion period, overseeing a $15 billion fundraise. He sat on the Pentagon's Defense Business Board during his Uber tenure. His appointment as under-secretary signals the administration values speed and deal-making over diplomatic niceties in AI procurement.

The Clause That Broke the Deal

One Phrase, Two Interpretations

Amodei revealed in a staff memo, seen by the FT, that the negotiations came remarkably close to succeeding. The Pentagon offered to accept Anthropic's terms - the red lines on surveillance and autonomous weapons - if the company deleted "a specific phrase about 'analysis of bulk acquired data.'"

Amodei told staff that phrase "exactly matched the scenario we were most worried about." Bulk data analysis is the technical foundation of mass surveillance programs. Removing the restriction on it would, in Anthropic's view, leave the other red lines unenforceable.

The Pentagon's position is that existing law already prohibits warrantless surveillance of Americans, making the clause redundant. Anthropic's position is that legal prohibitions and technical safeguards are different things - and that the Pentagon specifically asked to remove the one clause most relevant to domestic surveillance.

The Legal Fragility

The supply chain risk designation itself is on shaky legal ground. Lawfare's analysis, published this week, concluded the designation "won't survive first contact with the legal system." The problems are many: no notice was given to Anthropic, no opportunity to respond, the statute doesn't grant authority to bar private companies from doing business with each other, and the designation has historically been reserved for foreign adversaries.

Anthropic has already filed a lawsuit challenging the designation. Legal experts from Mayer Brown, the Center for American Progress, and Defense One have all published analyses questioning the Pentagon's legal authority.

The White House Complication

Axios reported Tuesday that the White House has cast doubt on whether a Pentagon-Anthropic reconciliation is even possible. An administration official told Axios: "Ultimately this is about our warfighters having the best tools to win a fight and you can't trust Claude isn't secretly carrying out Dario's agenda in a classified setting."

The administration's position appears divided: the Pentagon wants to resolve the dispute and restore access to what was previously its preferred frontier AI model, while the White House may prefer to make an example of Anthropic. Amodei's leaked internal memo - where he attributed the conflict to Anthropic not giving "dictator-style praise" to Trump and not donating to his campaigns - is seen as having made reconciliation harder.

This matters because the original crisis began with a Truth Social post from President Trump directing agencies to stop using Anthropic - not with a Pentagon procurement decision. If the White House doesn't back a deal, Michael's negotiations may be dead on arrival.

Stakeholder	Position	Timeline
Anthropic	Red lines on surveillance and autonomous weapons are non-negotiable, but willing to negotiate language	Active talks
Pentagon (Michael)	Wants Claude back, willing to negotiate terms	Active talks
White House	Has cast doubt on reconciliation	Unclear
Courts	Anthropic lawsuit pending, legal scholars skeptical of designation	Weeks to months
Defense contractors	Actively purging Claude from operations	6-month wind-down

What Happens Next

Three paths are open.

Path one: Amodei and Michael reach a compromise. Anthropic accepts narrower language on bulk data analysis that satisfies its surveillance concerns without the specific phrase the Pentagon objected to. The supply chain risk designation is quietly withdrawn. This is the path both sides appear to want.

Path two: The White House blocks reconciliation. The lawsuit proceeds. Courts rule on the legality of the supply chain risk designation. This takes months and leaves the Pentagon without access to Claude for now, while defense contractors complete their migration to alternative models.

Path three: The deal collapses permanently. Anthropic pivots fully to commercial markets, where it's already seeing record growth - Claude hit #1 on the App Store after the ban, and daily signups have broken records every day since. The Pentagon relies on OpenAI and Google for frontier AI. The precedent is set: AI companies that impose conditions on military use get cut out.

The market is watching path one. The administration may be choosing path three.

Sources: