How researchers broke ChatGPT and what it could mean for future AI development

Researchers bypassed the safety guardrails for ChatGPT, Bard, and Claude with a series of adversarial attacks.