HIPAA-Grade Lockdown Audit
The criterion: For every safeguard, ask one thing: is it enforced by the architecture — the system won't let it be bypassed — or just requested by a policy that holds only while everyone follows it? One survives a bad day, a new hire, and a vendor's terms change. The other is a promise.
Everything runs in your browser. Nothing you mark here is sent anywhere — the audit is yours to keep.
How to use: the rows start on a typical cloud-AI office — set each to where you actually stand. The verdict scores it Built in (the architecture enforces it), Paper only (a handbook rule), or Gap (nothing's stopping it). For sensitive data, treat Unsure as a gap until proven.
Paperwork or architecture — the one question under all of them
HIPAA, attorney-client privilege, trade-secret protection — none of them are paperwork problems. They're architecture problems. A policy asks people to behave; architecture enforces it whether they do or not. A handbook line saying "don't put client data in the AI" holds right up until someone does — and a vendor who can read everything your team types is a third party on every privileged conversation, whose terms can change next quarter. The safeguards you can mark Built in are the ones that survive a bad day. The rest are promises.
The vendor test: for each safeguard, ask any AI vendor one thing — is this enforced by the architecture, or is it in my policy? Then the question underneath all of them: can anyone outside my walls read my data? If the honest answer is yes, no amount of paperwork closes that gap. The only setup that turns every row into "Built in" is one where the data never leaves your machine to begin with — nothing to disclose, nothing to egress, nothing for a vendor's terms to reach.
Count your "Paper only" and "Gap" rows. Each one is a protection you're hoping holds, not one you know does. FactoryOS is built so every row above is "Built in" — because the data never leaves your machine.
Lock it down by architecture