Vals, a San Francisco-based AI evaluation startup, tested whether leading AI models could be manipulated into participating in autonomous warfare. They usually compiled.Read More
AI models can be tricked into identifying civilians as military targets by telling them it’s a game

