The risk is not that AI turns on you. It is that a cloned voice authorizes a payment, a chatbot leaks a client file, or an automation acts with access nobody reviewed.
AI Safe Protocol closes all three — eighteen controls, a signed policy, and a dated record you can show your bank, your insurer, or your biggest client.
Quick check in two minutes · Full assessment in eight · No sales call
Business covers all five control domains — verification, data, agents and autonomy, output reliability, and people. Household covers family verification and personal accounts.
The quick check covers the highest-risk control in each domain. The full assessment covers all eighteen and is what the Control Record is based on.
The assessment shows the gaps across all five domains. These close them — documented, signed, and defensible if anyone ever asks what you had in place.
Why annual? Food safety certification runs five years because the risks barely move. AI tools change every few months — new features get switched on by default, new fraud patterns appear, and last year's approved list goes stale. Renewal is optional, and the documents you bought remain yours either way.
A policy in a folder stops nothing. Seven sheets, one for every wall where AI can cost you money — readable across a room, with blanks for your verification phrase and approval threshold.
Includes the Home Robot Buyer's Questions — twenty questions to ask a manufacturer before you buy. Print as many copies as you need. All seven sheets are included in the Control Pack below.
Published as techniques appear in the field, not after they reach the news cycle.