Cryptographers Show That AI Protections Will Always Have Holes | Dark Hacker News