How it works

One engine. Four steps. Your own network.

Every request goes through the same four steps, in every sector. Only the detector changes.

What Suvard is

A control and evidence layer that runs on your own network.

It sits in the path of the AI traffic, between staff and systems on one side and the AI models on the other. For every AI request it does four things.

Finds

names, CPR numbers, addresses, account numbers and whatever must never leave the building, including what no list knows. The detector is trained for the sector; in healthcare a Danish clinical language model.

Masks

the findings with placeholders before the AI sees them, and shows the user what was found. No new workflow.

Stops

what must not happen, by your own rules for who may send what to which model.

Seals

every event in a signed chain where later changes can be detected, and which can be verified without trusting us.

Where does it sit?

Between your people and the models. Never instead of either.

Your side
Staff and line-of-business systems

Chat, drafts, decision support, agents. The workflow does not change; the user only sees what was masked.

Suvard · on your network
Control and evidence layer

Finds, masks, stops, seals. The keys are yours. No content leaves the building, and nothing is sent to us.

The models
Cloud AI or local models

ChatGPT, Claude, Gemini or a model in your own data centre. We never supply the language model itself.

An example from healthcare

What the employee sends. What leaves your network.

The same question to the AI, but without anything that can identify the patient. The user sees the findings and approves the placeholders before anything is sent.

What the clinician writes

Draft a discharge summary for Jens Hansen, CPR 300290-1234, Vestergade 12, Aalborg, admitted with atrial fibrillation, treated with Marevan.

Marked: what the detector finds. Example data is fictional.
What reaches the model

Draft a discharge summary for [NAME], CPR [CPR][ADDRESS], admitted with [DIAGNOSIS], treated with [MEDICATION].

The answer comes back with the placeholders in place; the content is re-inserted on your side.

The healthcare detector is a Danish clinical language model, trained without a single patient record, plus CPR number checks. In legal it is the same four steps with a detector for case documents, which is on its way. Finance, energy, defence, space and the public sector follow the same pattern later.

The evidence

A trail that can be verified without trusting us.

What a sealed trail looks like.

Each line binds what the AI delivered, what the human did, and a digital fingerprint of the outcome into a single seal. The keys are yours. The content never leaves the premises. If anyone alters a line afterward, the seal breaks, and anyone can verify this using a small, standalone utility that requires zero trust in us.

Illustration. The events are examples, not real data.

Hash-chained

Each event points to the previous one using a cryptographic fingerprint. If a single line is deleted or altered, the chain breaks, and the tampering is immediately evident.

Signed with your keys

The seal is signed using keys held by you, not by us. What is sealed is the proof of what occurred, never the content itself.

Open verifier

A small, standalone utility with no dependencies can verify the seal offline. Anyone can run it. That is what makes the independence more than a promise.

The same seal can be applied to non-conversational AI, such as draft entries in the EHR or decision support within the line-of-business system binding the AI's output and the human action into a single event. That is the evidence layer; fully built and currently piloting.

Why now

The law demands the evidence. The professionals say the infrastructure is lacking.

The AI Act already requires everyone who uses high-risk AI to be able to document how it worked and how it was used. In healthcare, the MDR adds to that. Neither defines an independent format for what the human did.

AI Act, art. 72(2)

“…actively and systematically collect, document and analyse relevant data which may be provided by deployers … on the performance of high-risk AI systems throughout their lifetime …”

The provider must continuously collect data on how the system performs in practice, also from those who use it. How that is captured, the law does not say.

AI Act, art. 26(6)

“… keep the logs … for a period appropriate to the intended purpose of the high-risk AI system, of at least six months …”

The duty to keep logs lies with the one who uses the system. A shared, verifiable format does not exist.

In healthcare also: MDR 2017/745, Annex XIV Part B

“… proactively collect and evaluate clinical data from the use in or on humans of a device which bears the CE marking …”

The manufacturer of CE-marked equipment must monitor its use in the clinic. One cannot follow up on advice if one does not know whether it was followed.

… clearly established in principle by active current regulations, but not enacted upon, given unclear standardization and requirements at the moment.
European Society of Radiology, consensus recommendations, December 2025An example from healthcare: the physicians' own society on monitoring AI devices in use. The requirement exists. The standard for meeting it does not.

Everyone demands the result. Nobody defines the format. That is what we build: an independent seal on top of the standards that already exist.

2 December 2027

The EU's high-risk requirements apply to Annex III: e.g. acute triage in healthcare, credit scoring of individuals in finance, and the management of critical infrastructure like energy.

2 August 2028

The requirements apply to AI that is part of CE-marked products (Annex I), e.g. medical devices, where most clinical decision support lives.

The dates cannot be moved. What can be changed is whether a format and references exist on the day they bite. Evidence cannot be produced retroactively.

Quotes: AI Act (EU) 2024/1689; MDR (EU) 2017/745; Digital Omnibus (EU) 2026/1744 (dates); Cuocolo et al., Insights into Imaging (2025). Not legal advice.

Where are you?

Two tracks, one layer.

The question belongs when the budget is set, not when the regulator calls.

AI in use today, without control?
The gateway
  • Mapping of which AI services are used, and how much, without personal data.
  • Danish recognition of people and sensitive details before data reaches a model. In healthcare: patient identity with a clinical model.
  • Guided redaction: the user approves the placeholders.
  • A signed log for every call, on your own network.
Validated in Treat Systems' technical environment
Ready for the requirements in 2027 and 2028?
The evidence layer
  • An independent seal on what your AI delivered and what the human did with it.
  • Also for AI without a chat interface, such as drafting case notes or clinical records, or decision support embedded directly in your core software.
  • Can be verified offline by others, without trusting us or the vendor.
  • No personal data to Suvard. The keys are yours.
Built, not yet in operation
The vendor ownsThe workflow

What the AI delivered and what the human did are recorded directly within the vendor's own interface using an open data model. We never build the user's graphical interface.

Suvard ownsThe seal

The event is sealed independently of the source system: hash-chained, cryptographically signed, with encryption keys held strictly by you. Never the underlying content - only tamper-evident proof of what occurred.

Open to allThe verifier

Anyone can run it and check the seal without asking us. That is what makes the independence verifiable instead of a promise.

Proof is only truly independent if the entity sealing it has no stake in the outcome. That is why Suvard operates entirely outside the operational workflow, and why we never supply the language model itself.

The rules behind

What Suvard helps you comply with.

GDPR

Data minimisation in practice: people's identity is masked before data reaches an AI model, and the user sees what was found.

NIS2

Danish law applies from July 1, 2025, covering healthcare, energy, finance, and public administration, among other sectors. The legislation mandates logging, monitoring, and that log data is protected against tampering. A signed chain where alterations can be detected provides precisely that.

AI Act (and MDR in healthcare)

Logs kept for at least six months by the deployer of the AI (Art. 26), and continuous operational performance data fed back to the provider (Art. 72 and MDR post-market surveillance). The evidence layer delivers both in a single audit trail.

Questions we hear

Straight answers.

Is Suvard a cloud service?

No. Suvard is installed on your own network, on your own infrastructure. Nothing sensitive leaves the building. That is the whole point.

Do we need to run AI models or buy GPUs ourselves?

No. Suvard sits in front of the AI services your staff already use. Local models are an option, never a requirement. We never supply the language model itself.

Which AI services does it work with?

Suvard sits in the path of the traffic as a proxy and is therefore vendor-independent: ChatGPT, Claude, Gemini and other services your staff use through Suvard's chat window or your systems call through Suvard. Copilot in Word, Outlook and Teams and AI built into a line-of-business system are not supported yet. Same engine, same evidence.

What is in operation today, honestly?

Healthcare, through CareProxy: the engine with Danish clinical detection, masking and a signed log is validated in Treat Systems' technical environment, and we are in dialogue with a Danish region. The evidence layer (the independent seal) is built but not yet in operation at a customer. The legal detector is on its way; detectors for finance, energy, defence, space and the public sector are not built yet. The engine is the same.

Does Suvard make us compliant with GDPR and the EU AI Act?

No tool can promise that. Suvard supports compliance: it enforces your rules technically and produces the evidence. The legal assessment remains yours.

What evidence do we get when the regulator calls?

A hash-chained, signed log of every AI request and every decision, which can be exported and verified with a small, standalone program, without trusting us.

How is this different from a firewall or classic DLP?

A firewall sees packets; classic DLP sees patterns. Suvard understands what is being sent to AI: it masks it with placeholders people can keep working with, stops what must never leave the building, and proves afterwards what happened.

How do we start?

A conversation, a mapping of your actual AI use, and then a pilot in your own environment. Write to matias@suvard.com.

Next step

A 30-minute check on where AI is already running in your organization.

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.