Is your assistant high-risk?

Determining whether your assistant is high-risk is a per-assistant exercise. The same organization can have both high-risk and non-high-risk assistants depending on how each one is used.

The assessment follows four steps:

1. Prohibited use check: Certain uses are banned outright. If any apply, the assistant cannot be deployed.
2. High-risk area check: If the assistant operates in a regulated area (e.g. welfare, recruitment, education), it is presumed high-risk.
3. Exception check: If the use case falls in a high-risk area, you may be able to claim an exception if the assistant only performs a narrow, procedural task.
4. Transparency check: Separate from risk class, certain types of assistants carry transparency obligations regardless of whether they are high-risk.

Work through the steps below. Document your answers: they serve as compliance evidence, particularly if you claim an exception.

---

AI Act Screening Assistant

For a more conversational and interactive experience, the AI Act Screening Assistant is available in the Assistant Library. It guides you through the same decision tree, asks follow-up questions where needed, and produces a full documented assessment with legal references that you can save. It communicates in both Norwegian and English.

---

Step-by-step questionnaire

Select everything that applies to your assistant in each step. The questionnaire will guide you to a result and tell you which obligations apply, and what you have to do to comply with them.

1 Prohibited2 High-risk3 Exceptions4 Transparency

Step 1 — Prohibited use AI Act art. 5

Does your assistant do any of the following? Check all that apply. If yes to any, the assistant cannot be deployed as described.

• Manipulate or deceive people in ways that impair their ability to make free decisions, causing harm Art. 5.1a
• Exploit vulnerabilities related to age, disability, or socioeconomic situation to distort behavior Art. 5.1b
• Evaluate or classify people based on social behavior or personal traits in ways that lead to detrimental treatment Art. 5.1c
• Assess the risk of a person committing a crime based solely on personal traits or profiling Art. 5.1d
• Create or expand facial recognition databases through untargeted scraping of images Art. 5.1e
• Infer the emotions of individuals in workplace or educational settings (except for medical or safety purposes) Art. 5.1f
• Biometric categorization used to infer race, political opinions, trade union membership, religious beliefs, sexual orientation, etc. Art. 5.1g
• Real-time remote biometric identification in publicly accessible spaces for law enforcement purposes Art. 5.1h
• None of the above

Check