Tool · AI Act / 2024/1689

ROLE AND RISK OF AN AI TOOL

A walk-through of Articles 3, 5, 6 and 50. At the end you see the role, the risk category and the deadlines as amended by Regulation (EU) 2026/1744.

The classification is done for one tool, not for the company as a whole. With one tool you may be a deployer with minimal risk, with another the provider of a high-risk system. Go through the questions for a single tool and then, if needed, start again for the next one.

The tool runs entirely in your browser. Nothing is sent to a server, the result is free and nobody asks for your e-mail. The logic follows the AI Act as amended by Regulation (EU) 2026/1744, in force since 27 July 2026 — older calculators and articles still work with deadlines that no longer apply.

An indicative aid, not a legal assessment The output is an informative aid, not a legal assessment or a binding opinion. Classification under the AI Act depends on what the tool actually does, what its intended purpose is and in what environment it runs — and a handful of questions cannot capture that in full. Borderline cases have to be assessed individually. The authoritative text is the official wording of the regulation as amended by Regulation (EU) 2026/1744.
1AI system
2Role
3Prohibited practices
4Risk category
5Result

AI system

Role

Prohibited practices

Risk category

An indicative result. It does not replace a legal assessment and borderline cases have to be assessed individually.

What follows from it

    This is an indicative aid, not a legal assessment. Classification under the AI Act depends on what the tool actually does and what purpose it is intended for; borderline cases have to be assessed individually. The authoritative text is the official wording of Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, with links in the sources below the tool.

    Open the guide →

    CypherOn provides cybersecurity consulting services and is not a law firm. For legal opinions, turn to an attorney registered with the Czech Bar Association.

    Guide to the AI Act

    The timeline after the omnibus, the four risk levels, roles, Article 50, cybersecurity of high-risk systems, penalties and supervision in the Czech Republic.

    Open the guide

    I have more questions

    What the amended Article 4 means for training, what may go into ChatGPT and Copilot, when a deployer becomes a provider, who enforces this in the Czech Republic.

    Open the FAQ

    Overlap with the Czech Cybersecurity Act

    Under Act No. 264/2025 Coll. an AI tool is an asset and its supplier a supplier of an ICT service. The calculator determines whether the act applies to you.

    Open the Cybersecurity Act calculator

    Sources

    Every statement about an obligation or a deadline above carries an article number that can be looked up in these sources. The authoritative text is the official wording, not this tool.

    Regulation (EU) 2024/1689 — Artificial Intelligence Act ↗ The official text published on 12 July 2024. The source of the definitions in Article 3, the prohibitions in Article 5, the classification in Article 6, the obligations under Articles 16, 23, 24, 26 and 27, transparency in Article 50 and penalties in Article 99. The 2026 amendments are not reflected in this text. Regulation (EU) 2026/1744 — the digital omnibus on AI ↗ A regulation of 8 July 2026, published on 24 July 2026, in force since 27 July 2026. The source of the new paragraphs 1a to 1c in Article 6, the new definition of a safety component in Article 3(14), the softened Article 4, the new prohibitions in Article 5, the additions to Articles 25 and 42, the move of Regulation (EU) 2023/1230 on machinery from Section A to Section B of Annex I together with the related wording of Article 2(2), and the postponed deadlines in Article 113. Annex III to the AI Act — high-risk areas of use ↗ The complete list of the eight areas and their sub-points is at the end of the official text. The selection in step 4 is a condensed overview of it, not a substitute — for a specific deployment the text of the Annex decides. Commission guidelines on the definition of an AI system ↗ Non-binding guidelines of 6 February 2025 going through Article 3(1) element by element. The source of the point that software merely executing fixed rules is not an AI system. Draft Commission guidelines on the classification of high-risk AI systems ↗ A draft published on 19 May 2026 covering both routes under Article 6, with examples of use. The consultation closed on 23 July 2026 and the Commission has announced the final text by the end of 2026 — until it is adopted, it remains a draft. Commission guidelines on the transparency obligations under Article 50 ↗ An interpretation of Article 50 for providers, deployers and competent authorities. They are complemented by the voluntary code of practice on the transparency of AI-generated content of 10 June 2026. AI Act Service Desk (European Commission) ↗ The official Commission contact point for questions on the AI Act, including an interactive walk-through of the obligations by role. ČTÚ — Artificial intelligence (in Czech) ↗ The page of the Czech Telecommunication Office on the AI Act. It states itself that the institutional and penalty mechanisms require a Czech implementing act which, as of the validity date of this page, is not in the Collection of Laws. Ministry of Industry and Trade — draft act on artificial intelligence (in Czech) ↗ A press release of 26 September 2025 on the division of supervision between the Czech Telecommunication Office, the Czech National Bank, the Office for Personal Data Protection, the Czech Office for Standards, Metrology and Testing and the Public Defender of Rights. This is a draft, not an act in force.
    Content valid as of 8 August 2026

    Classifying the tools that are already running

    Classifying one tool is easy.
    Knowing about all of them is the harder part.

    Assessment of AI tools before deployment and of those running in the company without IT knowing, including the data that leave for them. Write to us with what you are dealing with and we will go through the scope.

    Write to us →