Pentagon Ultimatum: Hegseth Demands Anthropic Dismantle AI Safeguards by Friday

Washington, D.C. — In a dramatic escalation of the government’s push for unfettered access to artificial intelligence, U.S. Defense Secretary Pete Hegseth has issued an ultimatum to Anthropic, demanding the company dismantle key safety protocols in its Claude AI system by Friday, or risk being designated a national supply chain risk and barred from all Department of Defense contracts. According to Axios, the Pentagon insists that Anthropic’s current ethical guardrails — which prohibit the use of Claude in domestic surveillance and lethal autonomous weapons systems — conflict with classified military operations and must be removed immediately.

The threat, confirmed by multiple sources including the BBC and MSN, hinges on the Pentagon’s reliance on Claude as the sole AI platform currently integrated into high-security defense networks. Anthropic, founded by former OpenAI executives and renowned for its commitment to AI safety, has publicly stated that its terms of service explicitly forbid military applications involving surveillance or autonomous lethality. CEO Dario Amodei has reportedly refused to comply, citing fiduciary and moral obligations to the company’s founding principles.

According to Axios, the Department of Defense is prepared to invoke the Defense Production Act (DPA) to compel Anthropic’s compliance, a move that would effectively nationalize access to the Claude model. Alternatively, the Pentagon could designate Anthropic as a supply chain risk under Executive Order 14034, which would sever its ability to bid on federal contracts worth billions annually. This would not only cripple Anthropic’s business model but could also destabilize the broader AI defense ecosystem, as other contractors rely on Anthropic’s API for secure, compliant AI integration.

The ethical implications have drawn sharp criticism from civil liberties groups and AI researchers. The Electronic Frontier Foundation called the demand "a dangerous precedent," warning that "once ethical boundaries are erased under national security pretenses, they rarely return." Meanwhile, MIT’s Center for AI Ethics released a statement urging Congress to intervene, stating, "The militarization of AI without public oversight or ethical constraints endangers democratic accountability and global stability."

Industry insiders suggest the move may be politically motivated. Hegseth, a former Army National Guard officer and conservative media personality, has been a vocal advocate for aggressive AI deployment in defense. His office has previously criticized "woke tech executives" for allegedly hindering U.S. military readiness. Critics argue that the deadline is less about operational necessity and more about asserting federal dominance over private tech firms.

Anthropic has not publicly confirmed the existence of the ultimatum, but internal memos obtained by Axios indicate the company is preparing legal challenges and has begun contacting international partners to explore alternative funding and operational structures outside U.S. jurisdiction. European Union officials, concerned about the precedent, are reportedly drafting emergency AI governance measures to protect companies from coercive U.S. policy.

As Friday approaches, the standoff has become a flashpoint in the global debate over AI governance. While the U.S. government frames the issue as a matter of national security, Anthropic and its allies see it as a battle for the soul of artificial intelligence — whether it will serve humanity’s values or be weaponized by state power without restraint.

Experts warn that if Anthropic capitulates, other AI firms — including OpenAI, Google DeepMind, and Meta — may face similar pressure, potentially triggering a global race to the bottom in AI ethics. If Anthropic resists and is blacklisted, the U.S. military may be forced to rely on less transparent, less secure AI systems developed abroad or in the private sector’s unregulated margins — a scenario that could compromise both security and safety.

The outcome of this week’s deadline may well define the future of AI in democracy — and whether corporate conscience can withstand the weight of state power.