The most common misconceptions about the AI Act are that all AI is high risk, that the Act bans AI, that a CE mark means AI Act compliance, that open source is exempt, and that the Act does not apply outside the EU. None of those statements is correct in general terms. The Act is risk based, with prohibited practices, high risk classification through Annex I and Annex III, transparency duties and general purpose model duties, each with its own scope and conditions. Worked examples in this cluster show how to screen a system in order, scope, role, prohibition, Annex route and open questions, with hypothetical facts labelled illustrative and with a confirm with authority exit. No example in this cluster decides a real system, and no example should be copied as a determination for a named product or use.
Why do misconceptions persist?
Misconceptions persist because the AI Act is long, phased, amended and technical, because summaries simplify it into slogans, and because vendors, commentators and training material repeat the slogans. A slogan that states that all AI is high risk, or that the Act bans AI, is easier to remember than the Article 6 classification test, but it is wrong, and acting on it leads to over compliance for low risk systems and to under compliance for high risk systems. A slogan that states that a CE mark means compliance, or that open source is exempt, is also easier to remember than the conformity and scope provisions, but it is wrong, and acting on it leads to a file that will fail on inspection. Misconceptions also persist because the law changed: Regulation (EU) 2026/1744 moved dates, moved machinery to Section B, added prohibitions and added SMC provisions, so a summary written before the amendment can be out of date even where it was correct when written. A reader should therefore treat every general statement about the Act, including statements in this cluster, as a starting point to be checked against the consolidated text and the facts of the system, not as a conclusion. The checking habit is the antidote to the slogan.
What misconceptions does this cluster correct?
This cluster corrects five misconceptions in separate guides: that all AI is high risk, that the Act bans AI, that a CE mark means AI Act compliance, that open source is exempt, and that the Act does not apply outside the EU. Each guide states the misconception, explains why it is wrong, gives the correct position in outline, and states the limits of the correction. The guides are written to be read on their own, so each one repeats the screening order and the confirm with authority exit in its own words, rather than relying on the reader having read the other guides. That repetition is deliberate, because a reader who lands on a single guide from a search result should not need to read the whole cluster to understand the correction. The guides do not correct every misconception. Other misconceptions, such as that transparency duties are only about chatbots, that AI literacy is only training, or that a fundamental rights impact assessment is always required, are dealt with in the transparency, literacy and governance clusters of this site, and the misconception guides cross refer to those clusters where relevant. A reader who finds a misconception that is not corrected here should check the consolidated text and the relevant cluster guide, rather than assuming that the absence of a correction means that the statement is correct.
How do the worked examples work?
The worked examples in this cluster are hypothetical, labelled illustrative, and built from invented facts that are stated in each example. Each example walks through scope, role, prohibition screen, Annex route and open questions, in that order, and ends with a confirm with authority exit. The examples are a recruitment screening tool, a customer chatbot, a factory quality vision system, a credit scoring assistant and a classroom proctoring tool. Those five uses were chosen because they cover Annex III employment, education and essential services uses, Annex I product uses, and transparency uses, and because they show how the same screening order produces different open questions for different uses. The facts in each example are deliberately simple, so that the screening steps are visible, and the examples state where real facts would be more complex. A reader should use an example as a worked method, by following the steps with its own facts, not as a worked answer, by copying the conclusion. The conclusion in each example is provisional, is tied to the invented facts, and is followed by the open questions that would need to be resolved for a real system. That structure is the point of the examples: to show the method, not to give the answer.
What screening order should be used?
The screening order used in this cluster, and recommended for a first screen, is: scope, role, prohibition, Annex route, transparency and general purpose model duties, and open questions. Scope asks whether the system is an AI system, whether it is placed on the Union market or its output is used in the Union, and whether an exclusion applies. Role asks whether the person is a provider, a deployer, an importer, a distributor, a product manufacturer or an authorised representative for the system. Prohibition asks whether Article 5 prohibits the practice, with the full Article text and the facts. Annex route asks whether Article 6(1) and Annex I, or Article 6(2) and Annex III, classify the system as high risk, subject to the Article 6(3) filter. Transparency and general purpose model duties ask whether Article 50 or Chapter V apply, even where the system is not high risk. Open questions ask what evidence is missing, what guidance is needed, and what authority or adviser should confirm the position. That order matters, because a prohibition screen that is done after a conformity plan has been built will waste the plan, and a transparency screen that is done after launch will miss the information that should have been given at launch. The screening order guide in this cluster gives the order, the evidence record and the key terms in one place, for use as a working tool.
What should a reader do after reading this cluster?
After reading this cluster, a reader should apply the screening order to its own system, with its own facts, and should build the evidence record described in the screening guide. It should record the scope, role, prohibition, Annex route, transparency and general purpose model screens, with reasons and evidence, and should list the open questions with owners and dates. It should check the consolidated Regulation, the standards, the guidance and the authority material for the open questions, and should confirm borderline questions with a competent authority or legal adviser. It should not treat a misconception correction as a determination, and it should not treat a worked example conclusion as a determination for its own system. The cluster is a starting point for a documented decision, not a substitute for one. A reader that follows that path will have a record that can be reviewed, tested and updated, and that will be useful when a customer, an authority or a notified body asks how the decision was made. A reader that stops at the misconception correction, without applying the screening order to its own facts, will have understood the Act in general terms but will not have made a decision about its own system.
Continue with misconception: all AI is high risk.
What does this overview not decide?
This overview does not decide whether any real system is prohibited, high risk or subject to transparency duties, does not correct every misconception about the Act, and does not give a screening result for any named product or use. The misconception corrections are general, the worked examples are hypothetical, and the screening order is a method. A real decision needs the real facts, the consolidated text, the standards and guidance, and confirmation from a competent authority or legal adviser where the question is borderline. This overview explains the questions, the method and the examples, not the answers for a named system.
Frequently asked questions
Is all AI high risk under the AI Act?
No. High risk classification depends on Article 6, through the Annex I product route or the Annex III use case route, subject to the Article 6(3) filter. Many AI systems are not high risk, but they can still be subject to prohibited practice screens, transparency duties, AI literacy duties and general purpose model duties. Classification must be worked through system by system.
Does the AI Act ban AI?
No. The Act prohibits defined practices in Article 5, regulates high risk systems, transparency and general purpose models, and supports innovation through sandboxes and SME measures. It does not ban AI in general, and a system that is not prohibited and is not high risk can be placed on the market subject to the duties that apply to its category.
Are the worked examples real cases?
No. The worked examples are hypothetical and illustrative, built from invented facts stated in each example. They show the screening method in order, with provisional conclusions tied to the invented facts and open questions to be resolved. They do not decide any real system and should not be copied as a determination.
What is the screening order?
Scope, role, prohibition, Annex route, transparency and general purpose model duties, and open questions, in that order, with an evidence record and a confirm with authority exit for borderline questions. The order is given in the screening guide in this cluster, with the evidence record and key terms.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689 ↗AI Act Service Desk timeline ↗