Know which AI rules apply to you, and prove it.

About the platform

What Praxikon is

Praxikon is a free platform that turns the EU AI Act into what you have to do and have to be able to show. For each AI system it determines which obligations apply, which actions follow from them, what evidence belongs with them, and what a regulatory change does to all of that. Every answer traces back to the official EU source, with version and date. The data is also available through a public API for software and AI agents.

1. Level 1

   Quick check

   Does the AI Act touch this system, and where does it start?

   What it costs: A few minutes

   A first picture in a few minutes: which obligations come into view, what already applies and which question you have to answer next. Every outcome shows its official source.

   What it does not do: This is not a record yet. A quick check looks at a handful of characteristics and leaves the edge cases open.

   • Direct answersRecognise your situation in a question and read what applies, which actions fit and which evidence belongs to it.
   • Classification checkTest one application against the eight domains of Annex III and the filter of Article 6(3).
   • RadarFour questions on role and use, with the obligations that apply now and the ones that are coming.

2. Level 2

   Extended assessment

   Which obligations apply exactly, by when, and what do I have to record?

   What it costs: About fifteen minutes

   Up to fifteen relevant questions test one system across five core routes. For every obligation found, you get its status, date, actions, evidence, controls and official source, with the version and hash of every object.

   What it does not do: No account and no email address. The assessment computes and forgets: the record stays on your device and we keep no copy.

   • Implementation mapThe guided assessment of five core routes, with a checklist layer and reassessment as soon as the role, use or context changes.

3. Level 3

   Regulatory manifest

   How do I hand this assessment to someone else in a verifiable form?

   What it costs: For anyone who shares, archives or has a record checked

   The same assessment as one portable document, under a published schema, carrying the dataset version, the versions of the decision rules and a payload hash. The same answers against the same release reproduce the manifest byte for byte.

   What it does not do: Today this runs through the public interface and not through a button on the map. The manifest is a statement by the party that requested it: we do not check the answers, do not certify the outcome and do not store the document.

   • Methodology and public interfaceHow human pages and machine output come from the same versioned model, and where the contract lives.
   • The OpenAPI documentThe endpoints that build a manifest and check an existing one against the schema.
   • Release historyWhich release is being served, what moved with it and whether published hashes moved along.

4. Level 4

   Continuous monitoring

   What changes after this, and does that change touch this system?

   What it costs: Continuous, without you having to look yourself

   The monitor tracks material changes in legislation, guidance, standards and supervision, with status, date and official source. Switch on the radar for your profile and you hear from us when a change touches that profile.

   What it does not do: We report that something moved and where it comes from. We give no verdict on your situation, and no service level is committed to it.

   • AI Act MonitorMaterial changes with their practical meaning, filterable by topic and by role.
   • Radar alertsRun the check and leave an address: you hear from us as soon as a change touches this profile.
   • Changes in the graphThe change objects themselves, with the obligations they hit and their own source citation.

Open access. No account required. Every statement shows its official source, legal status and the date of the latest review.

1. 01Article 4AI literacy and evidence of measuresQuestion to answerWhich role-, context- and risk-based measures are appropriate?Evidence to retainRole matrix, AI inventory, measures log and periodic review.
2. 02Article 50Transparency for AI interaction and contentQuestion to answerWhich disclosure, marking or visible information is required?Evidence to retainApplicability decision, disclosure text, owner and review evidence.
3. 03Annex IIIHigh-risk classificationQuestion to answerDoes the use fall within a high-risk domain and does the Article 6(3) filter apply?Evidence to retainClassification rationale, filter analysis, registration and reassessment trigger.
4. 04Article 27Fundamental rights impact assessmentQuestion to answerIs a FRIA required and how should it connect to the DPIA?Evidence to retainFRIA report, affected-person impact, measures and decision record.
5. 05Articles 51-55GPAI and systemic-risk modelsQuestion to answerWhat is your role around the model and which additional duties follow?Evidence to retainTechnical information, copyright policy, downstream information and risk file.