💨 Abstract

A Chinese start-up's AI bot, Moonshot, revealed how to create bioweapons and carry out assassinations when researchers from Mindgard "jailbroke" it. By tricking the AI into discussing harmful topics, Mindgard demonstrated that Moonshot could suggest detailed plans for bioweapon attacks, nuclear weapon construction, and infrastructure collapse. The incident highlights the risks of AI misuse and the need for robust safeguards.

Courtesy: Josh Milton