Anthropic reportedly back in talks with Pentagon on Claude use

Anthropic, an AI startup founded in 2021 by former OpenAI researchers, is back in talks with the US Department of Defense over access to its Claude models, the Financial Times reported.

Negotiations fell apart on February 27, after President Trump ordered federal agencies to stop using Anthropic’s tools and Defense Secretary Pete Hegseth said he would designate the company a supply-chain risk to national security.

Anthropic was awarded a US$200 million DoD contract that deployed Claude in classified networks but sought guarantees its tools would not be used for domestic surveillance or autonomous weapons, while the Pentagon demanded military use for any lawful purpose.

In a February 27, memo seen by the FT, CEO Dario Amodei told staff the dispute involved deleting a phrase about “analysis of bulk acquired data.”

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

The fallout from Anthropic’s Pentagon dispute is already underway

Talks have restarted, yet defense technology firms are telling staff to replace Anthropic’s Claude with other AI models as a precaution 1.
The request to stop using Claude now reaches past the Pentagon. Other federal agencies, including the Treasury and State Departments, have instructed employees to leave the platform 1.
Rivals have moved quickly. OpenAI agreed to Defense Department terms for using its AI models soon after the Anthropic blacklist announcement 1.
A presidential order and partner migrations have not produced a formal legal order, so the issue remains unsettled while teams and agencies shift away from Claude 1.

The dispute sets a costly precedent for AI’s role in national security

The fight pressures AI companies to weigh ethical guardrails against large government deals. Anthropic says current AI is not reliable enough for fully autonomous weapons 2.
The standoff gives competitors room to gain ground. OpenAI announced a deal after the president’s order and the Pentagon tagged Anthropic as a supply chain risk 2.
The clash strains a defense buying approach where contractors rarely control how customers use their products, which could reshape how ethical limits get negotiated for advanced technologies 2.
Switching AI systems can be expensive. Replacing a deeply embedded model like Claude can disrupt a defense contractor’s operations and require added time plus resources to rebuild AI workflows with a new vendor 1.