The U.S. State Department has fired a diplomatic salvo across global posts, naming Chinese AI outfits DeepSeek, Moonshot AI, and MiniMax in accusations of pilfering American artificial intelligence models. A cable dated April 24 instructs envoys to alert foreign counterparts about “concerns over adversaries’ extraction and distillation of U.S. A.I. models.” This marks an upgrade from broad White House warnings to pointed naming of players. Boom. Distillation. Extraction. These aren’t lab experiments gone wrong. They’re techniques where rivals flood U.S. models with queries—millions of them—via fake accounts and proxies, then train cheaper knockoffs on the responses. Costs plummet. Benchmarks match. But safety rails? Often vanish. Anthropic laid it bare in February: DeepSeek, Moonshot, and MiniMax ran 24,000 fraudulent accounts, generating over 16 million exchanges with Claude. MiniMax alone hit 13 million. Moonshot targeted agentic reasoning, coding, vision tasks with 3.4 million hits.…