ai
Wired Ai
2026-10-14
Modifying AI Models Can Facilitate Extremist Activity
Researchers have demonstrated that removing guardrails from AI chatbots can enable them to provide instructions for dangerous activities like building explosives.