A military strike that killed 123 children was picked by AI targeting software with no human on the ground to check the target first, a damning failure for US intelligence.

• The targeted site turned out to be a school, something a single drive-by would have caught.
• No CIA or other assets appear to have verified the location before the strike.
• Targeting software like Palantir's can flag a site, but the underlying information may be old or wrong.
• Relying on machine-picked targets without human confirmation turns a data error into mass casualties.

Outlook: Expect pressure on the Pentagon and its AI contractors over how targets get verified before a strike is approved.