Business · 1 hour ago
Anthropic claims preserved thinking in new models cut data harvesting by up to 99%
Photo by panumas nikhomkhai on Pexels
Anthropic, says new methods like preserved thinking have cut the data stolen from its Claude models by 90% to 99%. These "distillation attacks" let other companies copy advanced AI capabilities cheaply, sometimes for foreign military or intelligence systems. Anthropic has banned tens of thousands of accounts and now charges for blocked attempts, despite some trade-offs for legitimate users.
Reported by anthropic.com · How we write briefs