This presentation will discuss two aspects of 6G Native AI support: AI4Net, using AI technology to improve network performance and Net4AI, providing AI as a service by the network as a whole. Recently 6G discussions on vision and enabling technologies have become hot topics in both industry and academia. AI for wireless has attracted even more research in the past few years. By now, it is well understood that AI can be useful to improve wireless network performance especially for tasks such as resource management, network planning and power saving where precise mathematical models are difficult to obtain. Standardization efforts are ongoing to facilitate these applications. Less conclusive is the application of AI in the physical layer design, even though we have seen various scenarios where AI can indeed improve the link level performance. Breakthrough research is needed in this area to warrant a radical new design. One the other hand, there are many usage scenarios being proposed for 6G, ranging from down to the earth 5G evolved use cases to over the moon (literally) space applications. In this talk, we will focus on AI as a service for 6G, in other words, 6G serves as a platform to provide distributed and collaborative AI learning and inference services. This is motivated by the proliferation of AI applications, data privacy concerns and regulations, as well as the need for real-time AI. The existing cloud base centralized AI learning model would not be scalable and sustainable, while on-device AI has limited computing capability and power. Distributed and collaborative AI is thus the key to large scale AI applications. A 6G network integrating sensing, communication, and computing would be a promising platform to support such distributed learning and inference AI services.