A chatbot almost sent US troops onto a Chinese ship
The AI risk debate has spent the week on model weights and pacing pledges. Today it produced something the argument kept promising and never showed: a real near-miss where an analyst's chatbot, not a model's inner life, put an armed boarding party on a launch path.
A US military intelligence report that claimed a Chinese ship in the Middle East was carrying components of a nuclear weapons program was built with the help of a chatbot — and was "entirely false," according to CNN, which spoke to four people familiar with the episode. The report circulated across the US military this spring, during the war with Iran, and set off planning for an intercept: sources told CNN armed service members were preparing to board the vessel and military aircraft were in the air. Officials only dug into the underlying reporting just before the operation was to begin, and found the material the ship carried had been misidentified by the chatbot an analyst had consulted. One source told CNN the episode "almost started a war" — a US operation against a Chinese vessel risked escalation between the two countries.
The mechanics are the part worth sitting with. The analyst queried the bot about intelligence reporting on the ship's manifest that originated with US Special Operations Command Pacific in Hawaii; the tool fused open-source material with secret signals intelligence and reached the wrong conclusion about the cargo, then the analyst used AI again to package the finding into a standard intelligence report of the kind military officials treat as authoritative. It is not clear whether the chatbot was commercial or a government product — one former senior official told CNN the internal tools are "mostly just copies of the commercial stuff wearing lipstick." US Special Operations Command Pacific and the Pentagon did not respond to a request for comment.
The adoption is decentralized by design: different corners of the military and intelligence community run different tools under different safety standards, with no shared rule for verifying what those tools produce. Targeting is the live edge — one source told CNN that AI in targeting is "definitely ramping up" with no real guidance on how a human in the loop is supposed to catch civilian casualties or fratricide. We covered the governance side of this last week — A hotline for rogue military AI, before Trump meets Xi — but that debate is about models going rogue. This incident is the more mundane and more likely failure: a person in a hurry trusting a system that answered confidently. As one source put it, "AI allows you to get to a bad idea faster."
Disney hired its first chief technology officer, and he comes from the AI company Disney once accused of ripping off its characters. Karandeep Anand, until now chief executive of Character.AI, joins on October 2 as senior executive vice president and CTO, reporting directly to CEO Josh D'Amaro. Disney sent Character.AI a cease-and-desist in September 2025 over its hosting of copyrighted Disney characters; Variety and the Wall Street Journal report that some of Character.AI's technical staff are expected to follow Anand to Burbank.
Read the hire as a signal about where the technology fight now sits. Disney is not buying a chatbot — it is buying the person who ran one, after spending two years of litigation and licensing on the question of who gets to generate a brand's own characters. The company's AI exposure runs through engineering and streaming, the mandate now attached to its first-ever CTO role: the studio that wrote the cease-and-desist decided it needs the capability in-house, with the terms it sets itself.
Does a hallucinated intelligence report belong on the same risk list as model misalignment — and should the military publish verification standards before expanding AI in targeting? Tell us in the comments.
