Anthropic called on rival AI developers to adopt defensive protections against model distillation attacks