Post
By Cory Doctorow (GPG 0xBF3D9110957E5F4C)
doctorow.pluralistic.net
did:web:pluralistic.net
Even if you don't trust Chinese models, you can extract their training through a process called distillation and transfer them to models you *do* trust:
https://www.anthropic.com/news/detecting-and-preventing-distillation-attacks
2026-09-16T08:03:20.907Z