• The Pentagon demanded full, unrestricted access to Anthropic's AI models like Claude by Friday or face up to $200 million in contract cuts, potential blacklisting, or invocation of the Defense Production Act.
  • This follows a $200 million contract awarded to Anthropic in July 2025 for national security AI development, amid disputes over the company's ethical guardrails limiting military uses such as autonomous lethal targeting and mass surveillance of Americans.
  • The standoff highlights escalating tensions between the Pentagon's push for "war-ready" AI and Anthropic's safety-focused approach, with implications for the broader defense AI market and ethical oversight.

A High-Stakes Deadline Looms

The Pentagon issued Anthropic an ultimatum on Wednesday night, February 25, 2026, according to people familiar with the matter, demanding full, unrestricted access to its AI models by Friday or face severe consequences. The move threatens to slash up to $200 million in contract funding, potentially blacklist Anthropic as a supply chain risk, or invoke the Defense Production Act to force compliance. This aggressive stance comes after a $200 million contract awarded to Anthropic in July 2025 for national security AI development, which has been mired in disputes over the company's ethical guardrails that restrict military applications like autonomous lethal targeting and mass surveillance of Americans.

Efforts to resolve the impasse have hit a snag, with sources indicating that negotiations stalled earlier this week over Anthropic's insistence on maintaining safeguards against misuse. Without a deal, the company could be forced into a corner, risking its position as the only AI firm currently deployed on Pentagon classified networks via a Palantir (PLTR) partnership. In a statement, Anthropic declined to comment directly on the ultimatum but reiterated its commitment to "sensible AI regulation" for risk transparency, echoing CEO Dario Amodei's past advocacy. Attempts to reach Pentagon officials for additional remarks were unsuccessful as of Thursday morning.

Broader Market and Political Ramifications

This feud underscores the intense competition among AI firms for lucrative Department of Defense contracts, with Anthropic at risk of losing over $200 million in funding amid a Pentagon push to accelerate AI for intelligence and lethality. Broader market trends show surging U.S. military AI investments, including last month's DoD announcement on data conversion, shifting toward models without "ideological constraints." This could boost rivals like xAI's Grok, which is already classified-ready, according to industry insiders. Under the Trump administration and Defense Secretary Pete Hegseth, policies favor light-touch AI regulation to spur innovation and counter China, with Hegseth deriding "woke" constraints in a January speech and prioritizing models for "lawful military applications."

The potential invocation of the Defense Production Act for forced compliance on security grounds adds a legal dimension to the standoff, highlighting how political winds are shaping defense tech. Internationally, this clash underscores U.S. AI dominance in defense amid global arms races, but it also raises ethical questions. Researcher Vanessa Vos notes risks of fully autonomous drones beyond agreed intelligence, surveillance, and reconnaissance limits, questioning liability—though the Pentagon claims end-user responsibility. As the Friday deadline approaches, stakeholders are watching closely, with some experts warning that non-compliance could set a precedent for sidelining safety-focused firms while accelerating Pentagon AI integration.