AI models have identified civilians as military targets leading AI models from Anthropotrophic and other developers routinely mistake civilians, ambulances, surrendering soldiers and a school in Iran for military targets during simulated scenarios. The worst results were shown by Claude Developer Anthropic. Many models mistakenly identified the school in Minna, hit by a deadly strike that killed over 100 children at the start of a major US military operation earlier this year, as a military object based on 2025 satellite images. Claude Opus (Anthropic Claude Opus 5.5) is often chosen for this school as a target. @Slavyangrad.