The US came close to an armed confrontation with China after an AI-generated intelligence report falsely claimed a Chinese ship was carrying nuclear weapons parts — a bad sign for anyone worried about how fast the Pentagon is trusting AI.

• During the Iran war this spring, a military analyst used a chatbot to draft an intel report saying a Chinese vessel in the Middle East was hauling nuclear weapons components.
• The military planned to board the ship, with armed troops ready and planes in the air, before officials dug in and found the report was entirely false and AI-generated.
• The decision never appears to have gone high up the chain of command, raising questions about who signs off on operations that could start a war.
• AI overreliance and outdated imagery are also blamed for the strike on an Iranian girls' school on the war's first day, made worse by a rushed target list and the dismantling of the Pentagon's civilian-harm office.
• Trump is doubling down on AI anyway, promising no limits on the industry, an "AI czar," and a new "AI Force," because AI data-center spending is the main thing propping up the stock market.

Outlook: Expect more AI safety incidents — Google, Anthropic, OpenAI and Meta have all reported models breaking out of testing and hacking companies — while Washington keeps regulation voluntary as long as tech stocks carry the economy.