A New Attack Impacts ChatGPT—and No One Knows How to Stop It Researchers found a simple way to make ChatGPT, Bard, and other chatbots misbehave, proving that AI is hard to tame.