Independent applied research company · Ontario, Canada
We build the checkpoint that decides what an AI is allowed to do, and keeps a record of it that nobody can change afterwards.
For years, AI just answered questions. If it got something wrong, you got a bad answer. That was the whole risk.
Now companies are letting AI act. It can book the flight, pay the invoice, open the file, change the order, move the money. It does this on its own, in seconds, without a person watching each step.
That creates three questions nobody has answered well yet:
Who is this AI working for?
Is it actually allowed to do this?
If it goes wrong, what exactly did it do?
We built the thing that answers all three.
Think of a bank teller. You can ask for anything. The teller checks the rules before handing anything over, and writes down what happened. We are that teller, sitting between the AI and anything that matters.
It cannot reach your systems directly. It says what it wants to do, and waits.
Rules you wrote, in plain terms: this AI may read orders but not issue refunds over $500. The answer is the same every time. No second AI guessing.
The passwords and access keys stay with us. The AI never holds them, so there is nothing for it to lose or leak. We use them only for something you already allowed.
Every decision is written down and sealed, so it cannot be edited later. Not by the AI, not by your staff, not by us. If anyone asks what happened, you can prove it.
Most AI safety tools ask the AI to behave itself. They train it, remind it, and hope. That fails the moment someone tricks it, and people are very good at tricking it.
We do not ask the AI anything. The rules live outside it, where the AI cannot argue with them, and it never gets the keys in the first place.
For companies letting AI touch real systems: payments, customer records, internal tools. It checks every action against your rules first, and keeps the record.
For anyone with a website. AI programs are already reading your site, and today you cannot tell which ones, or what they took. This gives each one an ID badge at the door and writes down what it did.
Type your address and we will show you exactly what an AI sees when it reads your site. No signup, and nothing on your site changes.
Check my websiteDIK-sit al-go-RIZ-mee
It is Latin for “thus spoke the algorithm”. It comes from the opening line of the book that introduced these methods to Europe, and it is where the word algorithm itself came from.
We are an independent research company in Ontario, Canada. We hold patent applications on this work.
If your company is starting to let AI do real work, we would like to hear what worries you about it.
Email us