A guide for small teams about to put real information into AI tools. I wrote it after reading my own notetaker's privacy policy, DPA and help center, and finding it took an hour just to line the three up.

The method: list what's actually running, find the exact plan, answer four questions per tool, sort what you find and give it an owner

The setting name and direct links for OpenAI, Anthropic, Google Gemini, Microsoft Copilot and Teams, Granola, Otter, Fathom and Read.ai

A chapter on starting a new AI project in the right order

A printable worksheet

Who it's for: founders and small mission-driven teams, especially before a new AI project touches real data.

Guide + skill (free): the 16-page guide as a PDF, plus an agent skill that runs the same audit in Claude Code.

Vendor pack ($49): everything above, plus the vendor table. What each plan retains, trains on and commits to, with the clause linked and the date it was checked. It comes installed in the skill and as a PDF.

Vendor terms were read on September 16, 2026. They change every few months, so every row carries its own date.