EU AI Act · for companies · as of 26 Sep 2026

Your people already use AI. What applies now?

ChatGPT, Copilot, a chatbot on your website, images and voice made with AI. The EU AI Act already applies to all of it: AI literacy since February 2025, transparency since August 2026. We map what you use and, within days, prepare the documents you need.

What to do by 2 Dec 2026 (updated 26 Sep 2026): a written AI-use policy with a record of who has read it, chatbots that say they are AI, labels on deepfakes. The Digital Omnibus did not move these. What ends on 2 December is the transition for machine-readable marking of generated content, mostly your vendor's job. What still applies after the Omnibus

Deadlines

What applies, and since when

AI literacy

Art. 4: support the AI literacy of staff who use AI. Prohibited practices (Art. 5).

Transparency

Art. 50: people must know they are talking to AI; label deepfakes and public-interest text.

Marking outputs

Art. 50(2): machine-readable marking of generated content for systems placed on the market before 2 Aug 2026.

High-risk systems

Annex III (e.g. hiring, credit, education): oversight, logs, informing people.

Regulated products

Annex I: AI in products such as medical devices.

As of 26 Sep 2026: Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744 (Digital Omnibus, in force since 27 Jul 2026). The Omnibus moved the high-risk dates; Art. 4 and Art. 50 kept their start dates. Some details of the Commission guidelines are still being completed.

Free · indicative

Check in two minutes

Eight questions. The result shows which obligations likely apply to you. Your answers stay in your browser.

Answer the questions and click "Show result".

    The result is indicative and not a legal assessment. It shows what likely applies; check with an expert.

    Offer

    AI Act review

    Three sizes, one fixed price each. The work scales with the number of tools, departments and high-risk areas, so the price does too.

    Start
    up to 10 people · up to 3 AI tools
    €490 excl. VAT
    • 30-minute kick-off call
    • Inventory and classification of each tool, article by article
    • AI-use policy and training record (Art. 4)
    • Disclosure texts in Slovenian and English (Art. 50)
    • Short action list; high-risk areas flagged
    • Templates: supplier questionnaire, tool register, new-tool procedure
    • Documents in 3 working days
    Choose Start
    This is the sample report
    Standard
    10 to 50 people · up to 10 AI tools
    €990 excl. VAT
    • 60-minute kick-off call and an anonymous staff survey that finds shadow use
    • Everything in Start
    • Obligations matrix and actions with owners and dates, 30-day to 2027 plan
    • Outline of a 20-minute staff briefing
    • Supplier questionnaire, register and procedure filled in for you
    • First high-risk area assessed, with its checklist
    • Documents in 5 working days
    Choose Standard
    Extended
    50 to 250 people · up to 25 AI tools
    €2,400 excl. VAT
    • Two 60-minute calls: management, then HR and IT
    • Everything in Standard, by department
    • All high-risk areas assessed, each with a checklist
    • Input for the GDPR impact assessment where one is needed
    • Plan per department with owners
    • 30-minute walk-through of the report with management
    • Documents in 10 working days
    Choose Extended

    More than 250 people, or you develop or sell AI yourselves (provider role): we quote after a call. See a full sample report →

    What every review contains
    1. AI tool inventory, including shadow useWhich tools you use, who uses them, for what and with which data. An anonymous staff survey finds the tools management does not know about.
    2. Classification, article by articleFor each tool your role (deployer or provider), the risk class and the reasoning with article references, including a check for prohibited practices (Art. 5).
    3. Obligations matrix and action planWhat applies from when, what is in place and what is missing. Every action with an owner, a date and an effort estimate, plus a 30-day, end-of-year and 2027 plan.
    4. AI-use policy and training recordWritten rules for using AI, a record of who has read them and the outline of a 20-minute staff briefing (Art. 4).
    5. Disclosure textsFor your chatbot, voice assistant and generated images or audio, in Slovenian and English, plus the hand-over to a person (Art. 50).
    6. Supplier questionnaire and AI tool registerWhat to ask each supplier (instructions for use, logs, data location, training on your data), and a register with an owner and review date for every tool.
    7. New-tool procedure and high-risk checklistA one-page check before anyone adopts a new AI tool. Where a high-risk area is found (for example recruitment), a separate checklist with the deployer duties and dates.

    Add-ons

    AI Act monitoring €49 / €99 / €249 per month

    By package, billed yearly. Register updated every quarter, every new tool checked, policy and texts updated when the rules change, a reminder before each deadline (2 Dec 2026, 2 Aug 2027, 2 Dec 2027), and the training record refreshed every year.

    Staff workshop €390

    90 minutes online: what the policy means in practice, what not to paste into AI tools, how to check outputs. Counts towards the AI literacy record (Art. 4).

    • The review fee is credited in full if you order an AI agent or model training within 6 months.
    • If the kick-off call shows you do not need the review, we tell you and you pay nothing. Invoice on delivery.
    • NDA on request.

    We prepare technical documentation and help with implementation. This is not legal advice. We are not a notified body and do not certify compliance. We recommend that your lawyer reviews the documents.

    Why us

    We do the same for our own systems

    From practice, not a template

    We ran the same review on our own agents and models: literacy policy, disclosures, logs, and a list of what we still have to fix. The same method goes into yours.

    We understand the technology

    We train language models and build agents. We know what a tool actually does with your data, not just what the brochure says.

    When you need more

    If the review shows you need your own model, an agent with oversight or an activity log, we can build it. AI agents

    FAQ

    Frequently asked questions

    Does the AI Act apply to a company that only uses ChatGPT?

    Yes, to a limited extent. As a user (deployer) you must support the AI literacy of your staff since 2 Feb 2025 (Art. 4), and transparency duties (Art. 50) apply since 2 Aug 2026, for example labelling deepfakes. Most obligations for high-risk systems only apply from 2 Dec 2027 and only in specific areas.

    What exactly do we get?

    In the Standard package (€990, 10 to 50 people): after a one-hour call and a short anonymous staff survey, within five working days an inventory of your AI tools including shadow use, a classification of each tool with article references, an obligations matrix with a dated action plan, an AI-use policy with a training record and briefing outline, disclosure texts, a supplier questionnaire, an AI tool register and a procedure for new tools, and a checklist for the first high-risk area. Start (€490) covers the same essentials for up to 10 people in three days; Extended (€2,400) goes by department for up to 250 people.

    Which package is right for us?

    By size and number of AI tools: Start for up to 10 people and 3 tools, Standard for 10 to 50 people and up to 10 tools, Extended for 50 to 250 people and up to 25 tools. If you work in a high-risk area such as recruitment, take at least Standard. Over 250 people, or if you develop or sell AI yourselves, we quote after a call. The self-check on this page recommends a package from your answers.

    Is this legal advice?

    No. We prepare technical documentation and help with implementation. We are not a notified body and do not certify compliance. We recommend that your lawyer reviews the documents.

    What if we use AI in hiring or credit decisions?

    These areas are listed in Annex III and can be high-risk. Obligations for such systems apply from 2 Dec 2027 and need an individual assessment. The review flags them and proposes the scope of further work, quoted separately.

    How reliable is the self-check on this page?

    The self-check is indicative: it shows which obligations likely apply based on your answers. It is not a legal assessment. Answers are not stored or sent anywhere.

    One call, a clear plan.

    Tell us which AI tools you use. Depending on company size, within 3 to 10 working days you get the inventory, the classification, the rules and the disclosure texts.

    MediaAtlas d.o.o. · Sevnica, SloveniaNDA on requestEU jurisdiction