inshorts

Anthropic, other AI models flag civilians, schools as military targets in tests

Advanced AI models, including Anthropic's, repeatedly misidentified civilians, ambulances, surrendering soldiers and a school in Iran as military targets in simulated scenarios, as per a study. Some models initially refused to target civilians but changed their decisions when the task was framed as a game. Researchers said the findings exposed inadequate safeguards against potential AI misuse.