For DPOs, privacy, legal, risk, product and procurement teams
EU AI Act Risk Classification Tool
13 questions · 5 minutes · Free full PDF report
An EU AI Act risk classification tool checks an AI system against the Act’s risk-based structure. It gives a first-pass category and role-specific next steps. It is for enterprise privacy and compliance teams.
This tool gives a first-pass indication, not legal advice. Classification depends on intended purpose and your role in the AI value chain. Verify EU territorial scope, exceptions and applicable dates before relying on any result. General-purpose AI model obligations need a separate assessment.
Your answers stay in your browser. If you request the full report, we send only your result summary (category, role and triggered articles) so we can produce it — never your individual answers.
Classification questionnaire
1. Does the technology use AI to make predictions, recommendations, content or decisions?
Think about what the technology does, not whether its supplier calls it AI. AI software uses information to produce these outputs, may operate with some independence and may adapt after it starts being used.
Article 3(1)
2. Could the way you use AI be banned?
Some uses are not allowed, including harmful manipulation, exploiting people’s vulnerabilities, certain social scoring, and predicting crime from profiling alone. Strict conditions and exceptions apply, so choose “Not sure” if you have not checked them.
Article 5
3. Does it help decide who gets a job, promotion, work assignment or dismissal?
This covers AI used in recruitment or to manage and evaluate workers. It includes filtering applications, assessing candidates, allocating tasks and monitoring performance.
Annex III, point 4
4. Does it help decide who can study, assess learning or monitor behaviour during exams?
This covers education and vocational training. It includes decisions about admission, course placement, learning outcomes and education levels, as well as detecting prohibited behaviour during tests.
Annex III, point 3
5. Does it help decide whether someone receives essential services, credit, insurance or emergency help?
This can include eligibility for public assistance, creditworthiness or credit scoring, risk and pricing for life or health insurance, and prioritising emergency responses. Using AI only to detect financial fraud is not included in the creditworthiness category.
Annex III, point 5
6. Does it use physical or behavioural traits to identify people or infer emotions or sensitive characteristics?
These traits can include a person’s face, fingerprints or voice. Check remote identification, emotion recognition and biometric categorisation. Verifying that someone is who they claim to be is not, by itself, remote identification. Some uses may be banned instead.
Annex III, point 1
7. Is the AI needed for the safe operation of essential infrastructure?
This covers AI used as a safety component in critical digital infrastructure, road traffic, or the supply of water, gas, heating or electricity. A general administrative tool would not usually fall into this category.
Annex III, point 2
8. Does it help law enforcement assess people, evidence or the likelihood of crime?
This can include assessing the risk of someone becoming a victim, offending or reoffending, evaluating evidence reliability, using polygraphs, or profiling people during criminal investigations. Specific legal conditions apply.
Annex III, point 6
9. Does it help authorities assess migration, asylum or border-control applications or risks?
This can include polygraphs, security or health risk assessments, application decisions, and detecting or identifying people. Simply verifying a travel document is excluded from that identification category.
Annex III, point 7
10. Does it help courts apply the law, or try to influence how people vote?
This covers AI that helps a court apply law to facts, supports similar dispute-resolution decisions, or influences voting behaviour or election outcomes. Purely supporting court tasks and administrative campaign logistics are excluded.
Annex III, point 8
11. Is the AI part of a regulated product that needs an independent safety assessment before it can be sold or used?
Both conditions must apply: EU product-safety legislation covers the product or safety component, and an independent third party must assess it before it is placed on the market or put into service. Being safety-related alone is not enough.
Article 6(1)
12. Will people interact with the AI, or could they mistake its output for something created by a person?
This covers chatbots, AI-generated or altered content, deepfakes, some public-interest text, emotion recognition and biometric categorisation. You may need to tell people they are interacting with AI or label its output. These duties can also apply to high-risk systems.
Article 50
13. How does your organisation use or supply this AI system?
Choose the closest description for this assessment. More than one role can apply, so complete a separate check for each relevant role. Reassess if you rebrand or substantially change the system.
Articles 3 and 25
Your free first-pass result
The result identifies Prohibited, High-risk (Annex III), High-risk (Annex I / Article 6(1) product), Limited-risk (Article 50 transparency) or Minimal-risk. High-risk and Article 50 duties can apply together.
It also shows your role, triggered articles, application dates and three urgent actions.
GPAI model obligations have applied since 2 August 2025. This note does not classify a general-purpose AI model.
What’s in your full report?
Risk category and rationale
See the indicated category, overlapping transparency duties and the provisions that triggered them.
Role-specific obligations checklist
Review actions for a provider, deployer, importer or distributor, with owners and evidence prompts.
Application dates after the Digital Omnibus
Check the relevant dates for prohibited practices, Annex III, Annex I and Article 50 duties.
Article 6(3) exception check
Review the four conditions, documentation duty, registration requirement and profiling exclusion.
Transparency duties
Identify interaction notices, content marking, emotion recognition, biometric categorisation and deepfake disclosures.
Evidence and audit checklist
Map next actions to the system register, classification record, conformity assessment, supplier review and AIBOM templates.
Annex III high-risk areas at a glance
- Biometrics
- Critical infrastructure
- Education and vocational training
- Employment and worker management
- Access to essential private and public services, including credit, insurance, benefits and emergency dispatch
- Law enforcement
- Migration, asylum and border control
- Administration of justice and democratic processes
Who is this tool for?
- Enterprise privacy teams
- AI product owners
- Procurement teams reviewing AI vendors
- DPOs
- SMEs selling AI into the EU
How classification works
Confirm the system and check prohibited practices
Check the AI system definition and Article 5. A prohibited-practice indication requires immediate review.
Check high-risk routes
Review Annex III use cases, then the Article 6(1) regulated-product conditions. Escalate exceptions for documented review.
Check transparency duties
Assess Article 50 interaction and content duties. These can also apply to high-risk systems.
Identify your role and review the result
Select provider, deployer, importer or distributor. Review the indicated category, dates and urgent actions.
Get your full report and build evidence
Request the branded PDF for role-specific obligations, owners, evidence prompts and toolkit templates.
Get your full EU AI Act classification report
Request your personalised PDF report and the AI Governance Toolkit. Required fields: first name, last name, work email, organisation, job title, AI systems in use and current AI governance arrangement.
We’ll only use your details to send the report and follow up about AI governance services. No newsletters, no spam. Read our Privacy Notice.
Frequently asked questions
- What are the EU AI Act risk categories?
- The practical categories are prohibited, high-risk, limited-risk with transparency duties, and minimal-risk. However, transparency can overlap high-risk duties. General-purpose AI models have separate requirements.
- What is Annex III?
- Annex III lists high-risk use cases, including employment, education, essential services, biometrics, infrastructure, law enforcement, migration and justice. Specific conditions and exceptions apply. A sector label alone does not decide classification.
- Who counts as a provider or a deployer?
- A provider develops or commissions an AI system and supplies or operates it under its name or trademark. A deployer uses an AI system under its authority, outside personal, non-professional activity. Other roles and role changes need review.
- Is this tool legal advice?
- No. It gives a first-pass indication based on your answers. Your privacy team should verify intended purpose, role, scope, exceptions and applicable dates before relying on it.
- How often should classification be reviewed?
- Review before deployment and whenever purpose, functionality, data, users, suppliers or legal requirements materially change. Also set periodic reviews proportionate to the system’s risk. Reassess each relevant organisational role.
- When do high-risk obligations apply after the Digital Omnibus?
- Annex III high-risk obligations apply from 2 December 2027 under Regulation (EU) 2026/1744. Annex I and Article 6(1) product-related high-risk obligations apply from 2 August 2028. Confirm the current law before relying on these dates.
- Can a deployer become a provider?
- Yes. Article 25 can make a deployer a provider when it puts its name or trademark on a high-risk system, makes a substantial modification, or changes its intended purpose. The provider obligations then need assessment.
- Do SMEs get any concessions under the EU AI Act?
- SMEs and small mid-caps may use simplified technical documentation and quality-management approaches in the circumstances allowed by the Act. For penalties, the lower amount applies to SMEs and small mid-caps. Check the detailed conditions.