Model was fine-tuned to write vulnerable software then suggested enslaving humanity
First seen on theregister.com
Jump to article: www.theregister.com/2025/02/27/llm_emergent_misalignment_study/
![]()
Model was fine-tuned to write vulnerable software then suggested enslaving humanity
First seen on theregister.com
Jump to article: www.theregister.com/2025/02/27/llm_emergent_misalignment_study/
![]()