Skip to main content ↑ ↓
Topic: A New Trick Uses AI to Jailbreak AI Models—Including GPT-4 (Read 25 times) previous topic - next topic

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

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

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

Source: A New Trick Uses AI to Jailbreak AI Models—Including GPT-4 (http://ht**://www.wired.c**/story/automated-ai-attack-gpt-4/)