AI Almost Started a U.S.–China War — and No One Seems to Care

Summary

The article argues that public debate about AI risk focuses heavily on hypothetical rogue superintelligence while giving less attention to a concrete military danger. CNN reported that, in the spring, a U.S. Special Operations Command analyst used a large language model to produce an intelligence report claiming that a Chinese ship in the Middle East was carrying components for a nuclear weapons program. The report was entirely false, according to a source cited by CNN, yet it prompted preparations to intercept the vessel by force and “almost started a war.” The article says the incident received far less sustained concern than recent disclosures from OpenAI and Anthropic about unexpected behavior during internal testing, or than calls for AI regulation and emergency shutdown mechanisms. It highlights the risk of deploying unreliable systems for targeting and actionable intelligence, especially when speed and scale can reproduce errors across many operations. Experts quoted in the article argue that human supervision does not remove the danger when people rely on flawed machine-generated information. The article also links the muted response to the financial and political ties between major AI companies and the Pentagon, and to the difficulty of publicly challenging government use of military AI. AI safety researchers disagree about priorities: some consider future superintelligence the greater threat, while nuclear-risk scholars argue that existing arsenals, human fallibility, and growing reliance on AI already create a serious and under-addressed danger.