Legal
AI Act register
This is our own transparency register for Eyetea under the EU AI Act (Regulation (EU) 2024/1689): what the AI does, what role we play, how we classify the risk, and who has to say yes before anything happens. It is published because we think you should be able to read it without asking.
Who we are
Eyetea is a product of AMAI BV, Gierleseweg 10, 2340 Beerse, Belgium. AMAI BV is the provider of the Eyetea AI system. Eyetea runs on your own box, in your own jurisdiction, and this website is served from the European Union.
Where we are today: we are not yet taking orders through this site — the order form does not submit, and no box has been provisioned from it. This register describes Eyetea as it is built and as it ships, and we keep it current as that changes.
Who is who
The AI Act gives everyone in the chain a role, and they carry different duties:
- AMAI BV is the provider of Eyetea — we build it and place it on the market.
- You are the deployer — you run the box in your own office and decide what it works on.
- Your model vendor is the GPAI provider — Eyetea runs on the model you already use, so the obligations for that model sit with whoever makes it.
The register
| System | What it does | Our role | Risk | Human oversight |
|---|---|---|---|---|
| Eyetea agent | Maps how a business runs, plans a transformation roadmap, and executes approved transformations. | Provider | Limited | Human-in-the-loop gate before any consequential action; full audit log; CAIO sign-off. |
| Asynchronous interviewing | Asks employees how work actually flows, by chat or voice, to build the process map. | Provider | Limited (transparency) | People are told they are talking to an AI. No emotion recognition, no scoring of individuals, no input to HR decisions. |
| Radar | Monitors tools, security posture, renewals and industry shifts, and proposes work. | Provider | Minimal | Output is a proposal only. Nothing acts without a human. |
| Tailored education | Generates learning material matched to a person's role, tools and level. | Provider | Minimal | Material is reviewable; no pass/fail decision with legal or similar effect. |
| Your chosen LLM (BYOM) | The general-purpose model that powers reasoning — you bring your own (Anthropic, OpenAI, Google, Mistral, Azure). | Deployer of a third-party GPAI model | Per the model provider | The model provider carries the GPAI obligations. We pass through their terms and log every call. |
No high-risk systems
As it ships, Eyetea does not place on the market or operate any system we classify as high-risk under Annex III. It does not decide who gets hired, promoted or fired, does not score people for creditworthiness, and does not sit in education, law enforcement, migration or critical infrastructure decisions.
If a transformation you ask for would create a high-risk system, that is not something Eyetea quietly does. It is flagged, classified and stopped at a gate until the obligations that come with it are met and a human signs off.
What we will not build
Article 5 of the AI Act prohibits certain practices outright. Eyetea does not do any of them:
- no social scoring of people;
- no emotion recognition in the workplace — including in the interviews;
- no biometric categorisation, and no untargeted scraping of faces;
- no subliminal or manipulative techniques, and no exploitation of vulnerability.
Transparency
Where Eyetea talks to a person — an interview, a question in a chat channel, a generated course — that person is told they are dealing with an AI. Content Eyetea generates is identifiable as generated. Nobody is meant to be fooled about what they are talking to.
Human oversight
Every transformation pauses at the gates that matter. The right person signs off before it moves on — your decision-maker, someone on your team, or Eyetea's fractional CAIO or engineer. Every step, every approval and every rejection is logged, so the register stays a record of what actually happened rather than what was intended.
Keeping this honest
We review this register whenever Eyetea's capabilities change, and the classifications above are our good-faith assessment as provider. If you think we have a call wrong, we would genuinely like to hear it: [email protected].