DayNews.ai

LLMs that adapt to new jailbreaks on their own

A new training framework helps LLMs adapt their own safety rules to counter novel jailbreak attacks without human oversight.