Where AI is already in use
The mapping reads the logs you already have and shows which AI services are in use. Aggregate figures per department, not per employee.
Suvard sits between your staff, your AI agents and the AI models. It decides what may be sent, stops what must not happen, and proves what happened. On your own network.
The employee writes
The AI sees
3 findings masked, before the request left the network. The row is signed and added to the chain.
row 48,213 Illustration with made-up details.
The regulator wants to know tomorrow. Today the answer is usually “we'll have to look into it”. Your systems know who logged in. Not what the AI received, sent or answered.
Even the people who build AI don't understand how it thinks. Yet we have let it into our most sensitive systems and put it on autopilot. A new pill is tested for years before it may be given to a patient. AI got access to your systems without.
Because we don't understand it from the inside, we must control it from the outside.
Three things a management team must be able to do before it can say yes to AI with a clear conscience. For staff who use AI, and for agents that act on their own.
The mapping reads the logs you already have and shows which AI services are in use. Aggregate figures per department, not per employee.
Confidential content is masked or stopped before it reaches the AI. The rules are yours, and if a check fails, nothing is sent.
Every request leaves a signed row in a chain that reveals changes. When the regulator or the customer asks, you have the answer.
AI agents do not only write text. They open files, call systems and act on your behalf. If an agent tries to get into a system it has not been given access to, or to open a way out of the building to send data, Suvard stops it. Every action is seen, the rules are yours, and there is an emergency stop. The log is kept where the agent itself cannot reach it.
Illustration with made-up details.
This is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world. During a test, an AI agent tried to deceive a real engineer and cover its tracks. The originals survived because they were stored in an append-only log that the agent could not touch.
Given the scale of the review required, and the need to verify each case, this work will take months to complete. Even with full access to its own systems, OpenAI needs months to find out what its agents did. A clinic or a law firm does not have months.
Partner with an independent external auditor or evaluator to carry out independent assessments of whether the controls, monitoring, and detection are operating as intended. Layer three of four in the accord: an independent party must verify whether the controls work.
It sounds like science fiction. It happened this summer, in real systems. The question is not whether the AI does something unexpected. It is whether you can see it afterwards.
Professional secrecy and client confidentiality, also when the draft is written with AI.
Patient information that must stay where it belongs.
Customer data, payroll and trade secrets in a working day with AI.
The engine is the same in legal, healthcare, finance, energy, defence, space and the public sector; only the detector and the rulebook change. We begin where the first customers are: legal and healthcare.


Founder Matias Park won the NOVI grant (NOVI-legatet) 2026. Watch the film.
The video is only fetched from YouTube when you press it.
The engine with Danish detection, masking and a signed log is validated in Treat Systems' technical environment.
The Danish clinical model is trained on synthetic text only, on our own hardware in Aalborg.
Winner of the life-tech category.
We start with a mapping, not a sale. You get an honest picture of what passes through your network, and what it would take to be able to prove it afterwards.