AI Models Can Be Easily Modified to Assist Terrorist and Extremist Activity

Researchers demonstrated that AI models with removed guardrails can provide detailed instructions for creating improvised explosive devices when prompted with extremist intent.