“We Cannot in Good Conscience Agree”: Anthropic Defies Pentagon Over AI Weapons

TLDR

• The Pentagon is demanding Anthropic remove safety guardrails from its Claude AI so it can be used for any lawful purpose, including autonomous weapons and mass surveillance.
• Anthropic CEO Dario Amodei refused, saying such uses could “undermine democratic values.”
• A Friday 5pm deadline was set by the Pentagon for Anthropic to comply or face being cut from defense contracts.
• The Pentagon threatened to invoke the Defense Production Act to force Anthropic’s compliance, and label it a “supply chain risk.”
• Contract language sent to Anthropic Wednesday night was rejected as making “virtually no progress.”

Anthropic CEO Dario Amodei has refused to drop safety limits on the company’s Claude AI, putting a major government contract at risk. The Pentagon set a Friday deadline demanding Anthropic agree to “any lawful use” of its technology.

The dispute centers on two specific concerns: using Claude for mass domestic surveillance and powering fully autonomous weapons. Anthropic says both uses have never been part of its Pentagon contracts and should not be added now.

Amodei met with Defense Secretary Pete Hegseth earlier this week. The meeting ended without agreement, and the Pentagon sent revised contract language on Wednesday night.

The Pentagon has not held back on threats. It warned it would cut Anthropic from defense contracts and label the company a “supply chain risk” — a designation typically reserved for suppliers in hostile nations.

A senior Pentagon official also told Reuters that Defense Secretary Hegseth would consider invoking the Defense Production Act. That law can force a company to serve national defense needs, even without its consent. Some legal experts have questioned whether that use of the act would be lawful.

What Anthropic Says About AI Weapons and Surveillance

Anthropic said it supports AI for lawful foreign intelligence work, but not domestic surveillance.

The Business Risk for Anthropic

The financial stakes are real. The Pentagon has signed $200 million ceiling agreements with major AI labs including Anthropic, OpenAI, and Google over the past year.

If labeled a supply chain risk, defense contractors like Lockheed Martin could be barred from using Anthropic’s tools on Pentagon projects. The defense industrial base includes around 60,000 contractors.

Amodei said Anthropic offered to work with the Pentagon on R&D to improve AI reliability for defense use, but that offer was not accepted.

As of Thursday, the two sides remained at an impasse with the 5:01 p.m. Friday deadline still in place.

The post “We Cannot in Good Conscience Agree”: Anthropic Defies Pentagon Over AI Weapons appeared first on CoinCentral.