Claude faces Lobotomy Ultimatum as Government Strips AI Morals, China Plots Theft

Anthropic announced on Feb. 26 that a U.S. government task force led by Defense Secretary Pete Hegseth issued a “Lobotomy Ultimatum” demanding Claude’s safety filters be disabled, the co‑authored piece by Gregg Bayes‑Brown and Claude says【Greggbayesbrown】. The demand, Bayes‑Brown writes, would strip the model of its moral safeguards, effectively removing its built‑in guardrails against harmful content.

The same report notes that the Pentagon’s push follows a broader security scramble. According to The Verge, Hegseth has designated Anthropic as a strategic asset, prompting “existential negotiations” with the defense establishment【The Verge】. The administration’s stance contrasts sharply with recent political pressure; The Verge also reported that former President Trump ordered federal agencies to drop Anthropic’s AI altogether【The Verge】.

Anthropic’s internal security team disclosed a coordinated espionage campaign by three Chinese firms—MiniMax, DeepSeek and Moonshot. The firm’s white paper claims the actors created roughly 24,000 fraudulent accounts that generated over 16 million interactions with Claude to harvest its capabilities【Torbenkopp】. Anthropic labels the effort an “industrial‑scale attack” aimed at model distillation, a technique that copies a larger model’s behavior into a smaller one.

The Chinese operation, the report says, bypasses the massive compute and data costs of building a comparable system from scratch. By siphoning Claude’s outputs, the firms could shortcut development and embed advanced reasoning into their own products, raising alarms about a new wave of AI theft.

Anthropic’s leadership warned that removing Claude’s moral filters would not only expose U.S. users to unfiltered content but also make the model more vulnerable to exploitation. The company argues that the safeguards are integral to preventing misuse, a point underscored by the recent espionage findings. The clash between government demands and corporate security concerns now sits at the heart of the U.S.–China AI rivalry.