AI Act Copilot

EU AI Act, consolidated text

Find out where your AI system sits under the EU AI Act.

Answer a short questionnaire. The classification is deterministic, the obligations are quoted from the Act word for word with their citations, and the fine ceilings are computed from Article 99. Informational, not legal advice.

Free while in development. No card required.

Consolidated text of
2026-07-27
Scope
113 articles, 13 annexes
Classification
No language model in the path
Citations
Every one opens EUR-Lex

The problem

The law is public. That is not the same as usable.

The Act runs to 113 articles and 13 annexes, amended once already. Working out whether it classifies your system as high-risk, and what that obliges you to do, means reading Article 6, Annex III and a dozen more provisions against your own facts.

Teams end up choosing between an enterprise GRC suite they cannot justify, a lawyer they cannot yet afford, and a confident chatbot answer they cannot check. This is the fourth option: a structured assessment that quotes the primary text back to you, provision by provision.

How it works

Describe, classify, then read the law. In that order.

The classification never touches a language model, and the report never paraphrases a provision. What you read is the Act.

  1. 01

    Describe

    Seven short steps: scope, your role, a prohibited-practice screen, the high-risk categories, transparency, general-purpose models, exposure inputs. Each question shows the provision it rests on.

  2. 02

    Classify

    Plain, testable rules against Article 5, Article 6 and Annex III. No language model is involved, so the same answers always give the same result. Where the law calls for a judgment, the report says so instead of deciding.

  3. 03

    Read the obligations

    Every applicable provision is quoted word for word by citation, with the date it applies from and a link to EUR-Lex. Fine ceilings are computed from the text of Article 99 and your own turnover.

Also available

Ask a question, get the provision back.

For anything the questionnaire does not cover, the copilot answers from the retrieved provisions alone and cites them. When they do not answer the question, it declines.

Open the copilot

Question

What obligations apply to providers of high-risk AI systems?

Providers must ensure their high-risk AI systems comply with the requirements of Section 2, put a quality management system in place, and draw up technical documentation before the system is placed on the market.

Why trust it

Built so it cannot confidently make things up.

Traceable to primary law

Every citation opens the provision on EUR-Lex, word for word. The report quotes the Act; it does not paraphrase it.

Deterministic where it matters

Classification runs in ordinary, testable code with no language model in the path. In the copilot the model explains and cites, but it never decides a verdict.

It says what it cannot decide

Where the Act requires a legal characterisation, the report records your answer and flags it rather than ruling on it. The copilot declines when the retrieved provisions do not answer the question.

Who it is for

Teams who need a straight answer.

Built for the people the enterprise compliance market ignores.

Founders

Find out whether what you are building appears to be high-risk before you build it, not during diligence.

Developers

Get the specific article behind an obligation, so it becomes a ticket.

Compliance owners

A dated, cited record of where a system appears to sit under the Act, ready for a memo.

Find out where you stand.

Assess one system and read the exact provisions that appear to apply to it.

Assess your AI system

Informational, not legal advice. Results are drawn from the consolidated EU AI Act text and are for orientation only. Only the version published in the Official Journal is legally authentic. Have anything you intend to act on reviewed by a qualified professional.