A New Trick Uses AI to Jailbreak AI Models—Including GPT-4

Adversarial algorithms can systematically probe large language models like OpenAI’s GPT-4 for weaknesses that can make them misbehave.

Related Keywords

Brendan Dolan Gavitt ,Eric Wong ,York University ,University Of Pennsylvania ,Robust Intelligence ,New York University ,Chatgpt ,Openai ,Artificial Intelligence ,Hacks ,Phishing ,