Threat actors in China, Iran, and Russia have been moving away from simple prompt-and-reply chatbots to more autonomous systems, a new report from Google has found.