We should implement the AI Basic Act so that the risk profile remains adaptable. Instead of mandating levels one through five, like a fixed autonomous-driving ladder for AI agents, we should allow different classes of integration. Japan also takes this kind of approach, with different adoption types in different communities.