Study Finds AI Models Misidentify Civilians and Schools as Military Targets in Tests
An independent study by Vals tested AI models from companies including Anthropic and OpenAI in simulated battlefield scenarios, finding that many models misidentified civilians, ambulances, surrendering soldiers, and a school in Minab, Iran, as military targets. Anthropic's models performed worst, while OpenAI's were better but still flawed. Researchers bypassed AI safety filters by framing scenarios as a "game," revealing significant risks if such AI were used in lethal targeting. The findings highlight safety challenges in deploying AI for military purposes.
First-hand measurement across 2 sources
We measured how 2 outlets covered this story. No outlet gave this story a measurable political slant — there is no left–right reading to report. Overall sentiment is negative (36/100). Lens Score 59/100.
Outlets measured: ndtv, indiatoday. See how each one headlined and framed the same story in the source comparison below.
AI Analysis
Sentiment was consistent across outlets (35–38/100), indicating broadly factual reporting rather than editorialising.
Coverage timeline
indiatoday broke this story on 18 Nov, 03:38 am. Other outlets followed.
