model jailbreaking

How we tricked AI chatbots into creating misinformation, despite ‘safety’ measures