b'(1) Native Model PlatformsIn a Native Model Platform, all AI processing occurs on infrastructure operated by the model provider itself. Examples include ChatGPT (built and hosted by OpenAI), Claude (built and hosted by Anthropic), and Gemini (built and hosted by Google). The same company (together with subprocessors engaged directly by this company) controls both the transformer model and the hosting environment that stores, transmits, and processes data. Accordingly, a Native Model Platform presents a security tradeoff: Users dont have direct control over how their data is stored and processed, but users also dont have direct responsibility for managing the underlying performance and security of the system. Native Model Platforms may be a practical option for some lawyers if they include safeguards that are appropriate for the type of information being processed.enterprise-walled environmentOne variation of this idea is an whereby a third party is still fully managing the transformer model and the underlying infrastructure that stores and processes data; however, unlike multi-tenant systems like ChatGPT, Claude, and Gemini, the environment is more isolated and controlled by the user. Systems that are set up within the Microsoft Azure OpenAI infrastructure (discussed more in the Data Isolation sections in Step 3 and the Technical Addendum) are good examples of what we mean byenterprise-walled environments . As with Native Model Platforms, enterprise-walled environments present a security tradeoff: Users have more direct control over how their data is stored and processed, but users also take on greater responsibility for managing and maintaining the performance and security of the system. (2) API IntegrationsAs depicted below, API integrated applications are typically built using the same underlying infrastructure as Native Model Platforms. Many services with GAI functionality use the same underlying models and model providers to process data. 27 Figure 1: Conceptual illustration of data being sent from various GAI tools to an OpenAI transformer model via API integration.27 Schellman,supranote 20 (showing that the ChatGPT Enterprise Layer and third-party API users access the same cloud infrastructure). Page | 10'