Restricting public access to AI reasoning steps does more harm than good

From the articleAnthropic says new safeguards sharply reduce AI distillation attacks | ReutersAlso in Anthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation