Researchers find LLMs like ChatGPT output sensitive data even after it’s been ‘deleted’

According to the scientists, there’s no universal method by which data can be deleted from a pre-trained large language model.