Methodology and sources

Legal status as of 18 July 2026 · English edition 0.2.0 beta · shared assessment core 1.0.2

How the tool works

We compare your answers with a database of obligations, prohibitions and recommendations built from the wording of the AI Act and related EU documents. Every item in the report cites the exact article of the regulation with a link to the source. The assessment runs entirely in your browser: answers are stored only in its local storage (delete them with “Start over”) and are never sent to our servers — we cannot see them. Technical visit data (such as your IP address) is processed by the hosting provider (Vercel) under its own terms, as on any website. The English edition uses the same assessment core as the Czech original — identical answers produce identical results in both languages.

How the tool catalogue works

The catalogue path saves time: you select the tools you use (for example ChatGPT or Tidio) and confirm what you actually use them for. From the confirmed use — never from the product brand alone — the guide suggests answers to the same questions the manual path asks, explains each suggestion and shows everything for review and editing before the report is created. Your manual change always wins. A product brand by itself creates no obligation: the same tool can be used internally (minimal impact) or in ways the AI Act targets — the purpose, the people affected and the degree of human control decide. The tool list is indicative, says nothing about specific products’ features and can be extended with your own entry. Catalogue version 1.0.1, last reviewed 18 July 2026. Your tool selection, like your answers, is never sent anywhere.

Key dates

  • 2 February 2025 — prohibited AI practices (Article 5) and AI literacy (Article 4).
  • 2 August 2025 — rules for general-purpose AI models (GPAI models), governance, the penalty framework.
  • 2 August 2026 — transparency (Article 50): chatbots, disclosure of deep fakes and certain AI texts.
  • 2 December 2027 — target date of the approved delay for high-risk AI systems under Annex III (recruitment, scoring, education…); until the delay is published in the Official Journal, 2 August 2026 formally applies.
  • 2 August 2028 — target date of the approved delay for high-risk AI systems embedded in regulated products (Annex I).

Note: the delays were approved by the Council of the EU on 29 June 2026 in the “Digital Omnibus” package; the act was signed on 8 July 2026 and, as of this tool’s legal-status date, was awaiting publication in the Official Journal (it takes effect on the third day after publication). Until then the original dates formally apply. We monitor the status and update the content.

Who does the AI Act apply to (Article 2)?

The AI Act can also reach businesses established outside the EU: it applies, among other cases, where a provider places an AI system on the market or puts it into service in the EU, and where the output produced by an AI system is used in the EU. A provider develops an AI system and places it on the market or puts it into service under its own name; a deployer is a business using an AI system under its authority. This tool focuses on the deployer perspective and flags when you may also be a provider.

Who enforces the AI Act in the EU?

Supervision is shared: the European Commission’s AI Office oversees general-purpose AI models and coordinates consistent application, while national competent authorities designated by each member state supervise most other obligations and handle complaints. Several member states were still completing their national set-up as of the legal-status date. National law may add further requirements beyond the EU-level AI Act — this tool does not assess employment, consumer, data-protection or other national rules.

The Commission’s official tools

The European Commission runs the official AI Act Service Desk with a Compliance Checker. We recommend it as an authoritative reference. This tool differs in focus: plain language for small businesses, a prioritised action plan and editable Word templates. It does not replace the official resources.

Sources

What this tool does not do

  • It does not provide legal advice and does not replace an individual assessment — it is a qualified first orientation.
  • It does not cover in depth the obligations of AI providers (Articles 16–25; for general-purpose AI models Articles 51–55), importers and distributors (Articles 23–24), or the full deployer regime for high-risk systems (Article 26) — seek professional advice there.
  • It does not decide whether a specific system falls under the Article 6(3) exception — the result is a first orientation, not a classification of your system.
  • It covers the EU-level AI Act only. National employment, data-protection, consumer-protection and supervisory rules may add further requirements.
  • It never sends your answers to any server — so there is nobody who could read them; the hosting processes ordinary technical visit data only.

Changelog

0.2.0 beta (19 July 2026) — the English edition moved to its own domain, aiactwithoutpanic.eu. Old /en addresses redirect here permanently. Content and the shared assessment core 1.0.2 are unchanged.

0.1.2 beta (18 July 2026) — proper dark mode following the system preference across the whole tool and centred value cards on the landing page. Assessment logic, catalogue and documents unchanged. Shared assessment core 1.0.2.

0.1.1 beta (18 July 2026) — fixes from independent production QA: catalogue entry “Custom chatbot” fully localised (including search), English accessible name of the SME Union logo, a broader wording of the provider role option (first use in your own business, rebranding, substantial modification) and a non-wrapping language switcher on mobile. Shared assessment core 1.0.1.

0.1.0 beta (18 July 2026) — first public beta of the English EU Edition, built on the shared assessment core 1.0.0.