Anthropic detailed how Chinese AI companies attempt to reverse engineer LLMs like Claude using sophisticated distillation attacks.