Google, Apple, and others show large language models trained on public data expose personal information

Researchers at Google, OpenAI, and others show that large language models including GPT-2 can reveal sensitive information.