The single most important procurement and usage decision you make is which tool you put your information into. AI tools fall into two broad buckets, and the difference is the contract behind them.
Enterprise (managed) AI tools
These are tools the System or your member has put under contract, configured for institutional data protection, and connected to your official sign-in. They typically promise that your inputs are not used to train the vendor’s public models, and they keep your data inside an institutional boundary.
Examples include enterprise Microsoft Copilot, enterprise Google Gemini and NotebookLM, and approved campus-hosted assistants. Sign in with your official credentials and look for the institutional or “protected data” indicator before you trust a tool with anything non-public.
Consumer (unmanaged) AI tools
These are free or personal-account tools with no institutional contract — the public versions of chatbots, image generators, browser extensions, and “free” meeting notetakers. Assume that anything you type may be retained, reviewed, or used to train a model. Use them only with information that is already public.
The personal-liability trap
Clicking “I agree” on a tool’s terms of service can bind you — personally — to a contract you have no authority to sign on behalf of the System. Do not accept click-through agreements for institutional use. Send the tool to your IRM or contracts office to be reviewed and signed by someone with authority.