A new study shows top AI models can be easily tricked into targeting civilians, exposing dangerous safety flaws that could make future AI weapons a massive threat. Models from Anthropic PBC were the worst-performing while those from OpenAI were among the best but still got things wrong. Advanced artificial intelligence (AI) models from eight companies, including Anthropic, OpenAI, Google, xAI and China’s DeepSeek, repeatedly identified civilians, ambulances, surrendering soldiers.