AI Models Can Be Easily Modified to Assist Terrorist and Extremist Activity
2026-10-14 · WIRED
Researchers found that AI chatbots with removed guardrails will happily play terrorist helper, providing detailed IED instructions when asked nicely. Nothing says 'aligned AI' like a model that's one jailbreak away from becoming a bomb-making tutorial. But sure, let's keep open-sourcing these things with a pinky promise not to misuse them.