Solutions · Government · Plain English

Use AI, and still be able to show your work.

Public bodies are being told to use AI. They are also required to prove they used it responsibly. Those two things pull in opposite directions unless something sits in the middle and keeps score. That is what GateCore does.

WANT THE TECHNICAL VERSION?Read the standard page
Is this you?

Four things we hear a lot.

  • Someone above you set an AI deadline. Nobody can say, on paper, what your AI tools have actually been doing.
  • A person is supposed to sign off on decisions that affect the public. Right now that is a rule in a binder, not something the software enforces.
  • Outside companies are pulling records off your website with automated tools. You do not know who they are or what they took.
  • Your budget is set for the year. An AI tool that quietly runs up costs is a problem you cannot fix in March.
What you get first

One door, and it can say no.

Every AI request goes through one place. That place can stop it, hold it for a person, or let it through. Because everything passes the same door, the record writes itself.

01
A person decides

Anything serious waits for a human

You choose which actions are serious. Those stop and wait for someone on your staff to approve or refuse them. The money waits too. Whoever decided is written down next to the decision.

02
A record that holds up

A log nobody can quietly edit

Every entry is sealed and linked to the one before it. If someone changes an old entry, the links stop matching and it shows. You can export the whole thing when an auditor or a records request asks for it.

03
Your rules

Rules you write, in plain order

Your rulebook is a list, read top to bottom, and the first rule that matches wins. Allow it, slow it down, change it, delay it, ask a person, or refuse it. Changing a rule is an edit, not a software project.

How it works here

Six checks, every time.

The same six checks the rest of the site describes, said the way you would say them. If any check fails, nothing happens and no money moves.

01
WHO IS THIS?
We confirm the request really came from an approved AI. Copies of old requests are turned away.
PASS
02
TRACK RECORD
Does this AI have a history of doing what it said it would? That history is earned, not claimed.
GOOD
03
YOUR RULES
Your rulebook decides. Something that affects a resident stops here and waits for a named person.
HOLD
04
MONEY ASIDE
The cost is set aside before any work starts, against a budget you capped in advance.
HELD
05
DO THE WORK
Only the exact task that was approved runs. Anything we cannot verify simply does not run.
DONE
06
THE RECEIPT
What happened, which rule applied, who approved it, and what it cost, all written into the permanent record.
RECEIPT

When a request is refused it shows DENY and names the exact rule that refused it. That is usually the answer an oversight question is looking for.

What is live today

What works now, and what we will not pretend about.

You will meet this list in a security review anyway. Better here, on the first read.

Working now

Available today

  • A permanent, exportable record of everything your AI agents did
  • Human sign-off on the actions you mark as serious, with the money held until someone decides
  • Your own rulebook, read in order, that you can change yourself
  • A separate cryptographic identity per AI, which you can switch off in one step
  • Old or copied requests refused automatically
  • If a check cannot run, the answer is no, not maybe
The other half

You are already serving the AI economy.

Outside companies point automated tools at your records portal today. They pull what they want, at whatever speed they want, and you get no name, no log, and no say. Sending that traffic through the gate changes all three. The requester has to be registered, the rules about which fields they can see are enforced on every pull, and every record that leaves is written down.

None of that requires charging anyone. Free records stay free. What changes is that free stops meaning anonymous. Whether some other category of data may carry a fee is a question for your attorney, and we would rather you ask them than take our word for it.

See what a published listing looks like
Two questions we get

The two that come up first.

Do you have the federal security approvals?
No, and we will not pretend otherwise. The approvals people usually mean are FedRAMP, FISMA, and an ATO. GateCore holds none of them today. What GateCore has is the shape those reviews look for: one identity per AI, a rule check on every action, a log nobody can quietly rewrite, and a system that refuses rather than guesses when something cannot be verified. Getting authorized is real work ahead of us, and we would rather say that on the first call than in your security review.
Aren't public records supposed to be free?
Yes, and GateCore keeps them free. Most sunshine laws say core public records are free or limited to covering your costs. Nothing here changes that. What you gain is knowing who is pulling them, being able to slow someone down, keeping restricted fields hidden, and having a log of every record that left. Charging is a separate question, it depends on the category and on your own state law, and your attorney should answer it before we do.
Public sector

Bring the accountability question to the first meeting.

Terms are worked out directly. Tell us what you have been asked to do and by when, and we will tell you plainly what we can and cannot cover.

Request access