The standoff between the Pentagon and Anthropic over how artificial intelligence can be used in defense continued on Thursday as the A.I. start-up reiterated its reservations, a day before a deadline imposed by the Trump administration for the company to permit its powerful technology to be applied broadly for military operations.
The two sides are hurtling toward a 5:01 p.m. Friday deadline over a Pentagon demand that Anthropic provide unfettered access to its A.I. system without safeguards demanded by the company, as part of the negotiations over a $200 million contract involving A.I. in classified systems.
On Wednesday, the Pentagon gave Anthropic assurances that it would not use the company’s A.I. system, Claude, for mass surveillance of Americans or autonomous drone operations, which were the start-up’s key concerns.
But Anthropic said late Thursday that the new offer by the Pentagon fell short of what it was asking for.
“We cannot in good conscience accede to their request,” Dario Amodei, Anthropic’s chief executive, said in a statement about the Pentagon. “It is the department’s prerogative to select contractors most aligned with their vision. But given the substantial value that Anthropic’s technology provides to our armed forces, we hope they reconsider.”
If Anthropic does not comply, the Pentagon has said it will either invoke the Defense Procurement Act to force the company to work with it on its terms or designate it a supply chain threat and block the firm from doing business with the government. If the latter route is taken, the Pentagon and intelligence agencies could be forced to stop using Anthropic’s Claude, which has been the primary A.I. program used in classified systems.
On Thursday, Sean Parnell, the Pentagon spokesman, posted on social media that the Pentagon had “no interest” in using A.I. for the two scenarios Anthropic had expressed concerns about. But he reiterated that the Pentagon was demanding that the company allow the Pentagon to use the A.I. model “for all lawful purposes.”
“This is a simple, common-sense request that will prevent Anthropic from jeopardizing critical military operations and potentially putting our warfighters at risk,” he said. “We will not let ANY company dictate the terms regarding how we make operational decisions.”
Emelia Probasco, a senior fellow at Georgetown’s Center for Security and Emerging Technology, said an agreement between Anthropic and the Pentagon was critical. “Our operating forces need this deal to happen,” she said.
The fight began last month when the Pentagon asserted that Anthropic must allow its A.I. model to be used for any lawful purpose that the Defense Department wanted.
But Anthropic officials argued that its A.I. system was trained only to do certain things and would not be good at running drones or other tasks. The San Francisco company wanted assurances that its technology would not be used for the mass surveillance of Americans, provoking the Pentagon.
Anthropic had little time to review the latest offer from the Pentagon, a person briefed on the negotiations said. While the new language appeared to address Anthropic’s concerns, the new offer still gave the Defense Department the ability to set aside the restrictions when it saw fit, the person said.
Anthropic employees have been quietly cheering their leaders over the line the company has set with the Department of Defense, two people at the company said. Many employees were drawn to work at the company because it had positioned itself as valuing safety and more responsible A.I. technology, they said.
In his statement, Mr. Amodei said he believed deeply in using A.I. to defend the United States and other democracies. The company was the first to deploy an advanced A.I. model on classified systems, he said.
“Claude is extensively deployed across the Department of War and other national security agencies for mission-critical applications, such as intelligence analysis, modeling and simulation, operational planning, cyber operations and more,” he said.
Mr. Amodei said that while Anthropic had not raised objections to any particular military operations, some uses of A.I. were outside the limits of what “today’s technology can safely and reliably do.”
As a result, the company insisted on not allowing the use of the technology for domestic surveillance or autonomous drones, he said.
Julian E. Barnes covers the U.S. intelligence agencies and international security matters for The Times. He has written about security issues for more than two decades.
The post Anthropic Says It Cannot ‘Accede’ to Pentagon in Talks Over A.I. appeared first on New York Times.
