Anthropic says new safeguards sharply reduce AI distillation attacks | Reuters