A US Special Operations Command analyst used an AI chatbot this spring to synthesize open-source data with classified signals intelligence, producing a false report that claimed a Chinese vessel in the Middle East was carrying components for a nuclear weapons program. Airborne military aircraft were already en route and special forces units were preparing to board when a deeper review revealed the chatbot had hallucinated the cargo manifest entirely. The operation was aborted just before execution.
One source told CNN the intelligence was "entirely false" and "almost started a war." Officials could not independently confirm what the misidentified cargo actually was, nor whether the software used was a commercial product or a customized government tool.
Confirmed
- The false report originated with a Special Operations Command Pacific analyst in Hawaii who queried an AI chatbot to fuse open-source information with classified signals intelligence.
- The chatbot misidentified the ship's cargo as nuclear-weapons components; the analyst then used the tool a second time to format the erroneous findings into an official-looking summary that circulated across command channels.
- Military aircraft were airborne and boarding teams were readied before the error was caught; the intercept was called off at the last minute.
- The incident occurred during the ongoing conflict with Iran, adding operational tempo pressure on intelligence production.
- Defense Secretary Pete Hegseth's "AI Acceleration Strategy," mandated by President Donald Trump, aims to put frontier models into the hands of three million military and civilian personnel across all classification levels through a program called GenAI.mil.
Unknown
- The actual cargo on the Chinese vessel.
- Whether the chatbot was a commercial large language model (e.g., GPT-4, Claude) or a government-customized variant.
- What specific verification or human-in-the-loop safeguards existed at the time, and whether they have been updated since the incident.
- Whether any formal investigation or after-action review has been completed, and what disciplinary or procedural changes resulted.
- The exact date of the near-intercept — sources say "this spring" but no precise day has been disclosed.
Our take
The episode exposes a structural risk: when adoption speed outpaces verification infrastructure, hallucinations travel up the chain of command faster than reviewers can catch them. The Pentagon's push to deploy AI to three million personnel without uniform safety standards creates fragmentation—each service operates under distinct rules, and young analysts face pressure to produce intelligence rapidly. Until mandatory human-in-the-loop checkpoints are standardized, AI-assisted drafting will remain a flashpoint for escalation.
