The work is handed out
Each obligation gets an owner, a due date and the proof attached. Hand a department its obligations in one click, with a due date you choose.
LegisGate · the laws that apply
Which laws reach this AI use case, and the citation for each.
The determination is the first thing LegisGate makes for an AI use case. It names the laws that apply, the provision each duty comes from, and the laws that were checked and found not to apply. Every citation resolves to the official text.
Intake
About 15 minutes
Included
With every AI use case
Where it lives
On the use case, in your workspace
Inside a determination
Written to be handed to counsel without a covering explanation.
The test we design against: a careful lawyer opens it, pastes a citation into a memo, opens the source, and finds the provision says what we said it says.
01
The laws that apply
Named in full, with the pinpoint provision each duty comes from — not a nickname, not a product label, not a sibling article that happens to look statute-shaped.
02
The laws that do not
With the reason. A regulator asking why you did not run a particular assessment is answered by a dated record saying the law was checked and why it did not apply — not by the absence of a document.
03
What each law requires
The obligations, each carrying the citation it was determined from, sorted by the department that has to act on it.
04
Timing language that matches where you actually are
The intake asks whether the use case is not yet live, in pilot, in production, or in production and inherited from before this process existed. Deadlines and statutory timing sentences are written to that — because assess before deploying reads differently to somebody who went live in March. It does not soften which duties apply.
05
Language that matches what you told us
Because the intake asks which controls already exist, a finding can say if not already in place rather than asserting you have not done something. Where we did not ask, the determination does not claim to know.
06
Action items with a landing role
Pre-authored against the obligation and routed to a role that exists in an organisation like yours — not a model's guess at what you should do about a law it has just read.
07
The evidence you will be asked for
What to retain, tied to the obligation it proves. In the workspace the proof is attached to the obligation itself, so the two cannot drift apart.
08
The written assessments it calls for
Where a law requires a DPIA or a state privacy risk assessment, the determination says so, and the assessment is opened as a draft in your workspace when the determination runs. Where a law requires a FRIA, it is listed as waiting until its official template is published. How the written assessments work.
What this is not. LegisGate produces regulatory intelligence and assessment-preparation materials. Nothing we produce is legal advice, a legal opinion, a certification, or a determination of compliance, and no document we deliver satisfies a legal obligation on its own. Sufficiency is determined by your counsel.
How it is made
Deterministic where the law is deterministic.
Applicability is decided by code reading a curated corpus. A language model never decides which laws apply to you, never writes a finding, and never mints a citation.
Step 01
You describe the AI use case
The tool, what it is used for, the jurisdictions where the people are, the industry and the data it touches set the scope — then the intake asks what those turn on, and separately which controls you already have: the processing agreement, the transfer safeguard, retention rules, the named oversight owner, AI literacy training. The rest of it.
Step 02
Facts, not impressions
Your answers are recorded as coded answers, and some facts are derived from them by written rules. An unanswered question is never silently treated as a “no”. Where a law was kept in because a question went unanswered, the record says so.
Step 03
Every law is tested against its written conditions
Each law carries written conditions in the Meridian: where the people are, what kind of organization, what the use case does, what data it touches. Your coded answers are matched against those conditions by deterministic code, never by a model and never by reading prose. A law is in, or out with the condition that failed recorded. Each obligation is tested the same way.
Step 04
Duties come from the register, not from prose
Obligations are selected from a curated register tied to pinned provisions. Where the register cannot support a duty, the determination leaves it out rather than improvising a plausible one.
Step 05
Every citation is checked before you see it
Each citation is checked against the verified corpus before it is shown. One that does not resolve to a verified provision is held back, not printed.
What happens to it next
A determination is the start of the record, not a file to keep somewhere.
It sits on the AI use case in your workspace, and the rest of the platform works from it.
The drafts are opened
The DPIA and the state assessments the determination calls for are opened as working pages, with the regulatory half written and cited and your half marked as yours, for your counsel's review.
It is watched
When the law behind the use case changes, you are told which of your records the change touches.
Start
One AI use case.
About fifteen minutes of intake.
Pick the use case that would be hardest to explain in a letter. You will know which laws apply, which do not, and what each one asks of you.
Your first five AI use cases are free, with everything included: the laws, the obligations, the written assessments and monitoring. No card. More than five is a subscription.
The first one tells you the most
Add the hardest one first.
If the AI use case you are least comfortable explaining comes back clean, the rest of the inventory probably will too. If it does not, you have found the work — with a citation attached to it.