Quanta Magazine

Cryptographers Show That AI Protections Will Always Have Holes

Large language models such as ChatGPT come with filters to keep certain info from getting out. A new mathematical argument ...