Require an AI supplier to return the customer’s data and usable work products when requested or when the contract ends. A strong version covers logs, prompts, outputs, configurations, evaluations and documentation needed to rebuild workflows.
Agree in advance how an institution can stop using an AI supplier. Specify notice, transition assistance, data return and continuity so ending a contract does not also mean losing the ability to perform essential work.
Require compatible interfaces so an AI assistant can use tools and business systems without every connection being rebuilt for one supplier. Separate the assistant, execution environment and system of record where possible, while keeping permissions and action controls intact.
Require a dominant operating system to give competing AI assistants access to the device features and interfaces they need to work. Equivalent access can let users choose another assistant without losing essential functionality.
Let users choose which AI assistant is the default interface on a device or service and change that choice easily. This could prevent an operating-system or platform owner from making its own assistant the unavoidable gateway to everyday work.
Make it possible to move to another AI service without abandoning accumulated work. Data portability moves files and history; context portability preserves relevant preferences and projects; operational portability restores workflows, connections, permissions and other working arrangements.