The EU AI Act and your AI agents
What the Act asks.
Who provides and who deploys, which dates apply now that the AI Omnibus is in force, and which building blocks the platform gives you.
Checked on
The AI Omnibus has been in force since 27 July 2026. It moved the high-risk rules for Annex III uses from 2 August 2026 to 2 December 2027.
This page explains how the platform relates to the Act. It is not legal advice.
The dates
When each rule applies.
The timetable as it stands after the Omnibus.
2 February 2025
Prohibited practices
The practices the Act bans outright apply, along with the duty to build AI literacy (since simplified by the Omnibus).
2 August 2025
General-purpose AI models
Obligations for providers of general-purpose models, such as the model providers AI agents run on.
2 August 2026
Transparency
The transparency duties of Article 50, such as telling people they are dealing with AI where the Act requires it. These were not postponed.
2 December 2027
High-risk systems in Annex III
Postponed by the Omnibus from 2 August 2026. Covers uses such as employment and recruitment, credit scoring, education, critical infrastructure and biometrics.
2 August 2028
High-risk AI in regulated products
Annex I: AI that is part of a product already covered by EU product law.
Who is who
We provide the platform, you deploy it.
The Act gives duties to providers and to deployers. Infinized provides the platform; when you configure AI agents and put them to work, you are the deployer of what you built, and you decide whether your use is high-risk.
What we do
Build the controls into the platform, keep them working, host your data in the Netherlands under Dutch law, and tell you precisely what the platform can evidence when a procurement questionnaire asks.
What stays with you
The intended purpose of what you build, whether it is high-risk, and the duties that follow from that. Using AI in recruitment or credit decisions is where that question matters most.
Building blocks
What you build on.
What the Act asks, and the parts a deployer uses to meet the obligations that apply to them.
Human oversight
Approval gates on actions that change something, a named reviewer per kind of action, and checkpoints in flows where a person decides before the run continues.
ApprovalsRecord-keeping
An audit trail of every run, tool call, approval and setting change, with the cost and outcome of each run, exportable as a file and as OpenTelemetry traces to your own tooling.
Audit trailAccuracy and robustness
Strict grounding, so an AI agent answers only from its sources or says it cannot; evals that catch a prompt change before it ships; guards in front of risky tools.
SecurityTransparency
AI agents are named and marked as AI in the product, and every action they take is labelled in the transcript. Where people must be told they are dealing with AI, the platform gives you the means; deciding where that applies is yours.
ChannelsData governance
Knowledge scoped to the AI agents who need it, hosting in Dutch datacentres under Dutch law, and export and deletion of your data without opening a ticket.
Where it runs
Questions
What procurement asks.
Is Infinized AI Act compliant?
Compliance belongs to a use of AI, so the platform gives a deployer the building blocks the Act asks for (human oversight, record-keeping, accuracy, transparency), and we will tell you what it can evidence.
Are we a provider or a deployer?
Infinized provides the platform. When you configure AI agents and put them to work, you are the deployer of what you built. You can become a provider yourself, for example by putting a system on the market under your own name. Our AI usage terms set this out.
Does Infinized decide whether our use case is high-risk?
No. Whether a system is high-risk depends on its intended purpose and the Act’s annexes, and that classification is yours to make. Most of what runs on the platform is not high-risk; recruitment, credit decisions and education are examples of uses that can be.
When do the high-risk rules apply to us?
For the uses listed in Annex III, from 2 December 2027, after the AI Omnibus postponed them from 2 August 2026. For AI inside products covered by EU product law, from 2 August 2028. Prohibited practices and the transparency duties were not postponed.
What does the platform never do?
It does not build or help build a capability whose purpose is a practice the Act prohibits, and it will not classify your use case or take on your duties as a deployer.
Bring the questionnaire.
Send us the AI Act section of your procurement questionnaire. We answer it with what the platform does and what it can evidence, and we say plainly where a question is yours to answer.