Skip to content

Guides

EU AI Act risk check

Answer up to six questions about one use of AI and see which of the four risk levels of the EU AI Act it is likely to fall under, what that means for you and from when.

Written and reviewed by Solorai

In short

The EU AI Act sorts every use of AI into four levels: prohibited, high risk, transparency and minimal. This check walks one use through the tests in Articles 5, 6 and 50 and in Annex III, as amended by Regulation (EU) 2026/1744, and shows the likely level, what it means and the date it applies.

It gives a first indication for planning, not legal advice. Classify each use, not each product: run the check once for every way you use a system.

Check one AI use

Think of one concrete use, such as an assistant that screens job applications. Answer for that use only.

Does the use match a prohibited practice?

Article 5 bans these practices:

  • Manipulative or deceptive techniques that distort a person’s behaviour and cause significant harm
  • Exploiting vulnerabilities linked to age, disability or social or economic situation
  • Social scoring that leads to unrelated or disproportionate treatment
  • Predicting the risk that a person commits a crime from profiling or personality traits alone
  • Building facial recognition databases by untargeted scraping of images
  • Inferring emotions at work or in education, except for medical or safety reasons
  • Biometric categorisation that deduces race, political opinions, religion, trade union membership, sex life or sexual orientation
  • Real-time remote biometric identification in public spaces for law enforcement, outside narrow exceptions
  • From 2 December 2026: creating or manipulating intimate or sexually explicit images of identifiable people without their consent, or child sexual abuse material
Is the AI a safety component of a regulated product?

This covers AI that is a safety component of a product under the EU product laws listed in Annex I, such as machinery, medical devices, toys, lifts or vehicles, or that is itself such a product, where the product needs a third-party conformity assessment. AI used only for convenience, efficiency or quality control is not a safety component, unless its failure would endanger health or safety.

Is it used in one of these areas?

Annex III lists the high-risk areas:

  • Biometrics: remote identification, categorisation by sensitive attributes, emotion recognition
  • Critical infrastructure: safety components for digital infrastructure, road traffic, water, gas, heating or electricity
  • Education and training: admission, assessment of learning, monitoring during tests
  • Employment: recruitment and selection, decisions affecting the working relationship, task allocation, monitoring of performance or behaviour
  • Essential services: eligibility for public benefits, credit scoring, pricing of life and health insurance, triage of emergency calls
  • Law enforcement
  • Migration, asylum and border control
  • Justice and democratic processes: assisting courts, influencing elections
Does it profile people?

Profiling means automated processing of personal data to evaluate aspects of a person, such as performance at work, economic situation, health, preferences or behaviour.

Does it only do a limited task?

Under Article 6(3), a use in an Annex III area is not high-risk when it poses no significant risk and does not materially influence the decision, because it only:

  • Performs a narrow procedural task
  • Improves the result of a human activity that is already completed
  • Detects patterns in past decisions, without replacing or influencing the human assessment
  • Performs a preparatory task for an assessment
Does it interact with people or generate content?

Article 50 covers systems that:

  • Interact directly with people, such as chatbots and voice agents
  • Generate text, images, audio or video
  • Create deepfakes

Likely prohibited

The use matches a practice banned by Article 5. Stop using it. The bans have applied since 2 February 2025, and the two added by the Digital Omnibus apply from 2 December 2026. Fines reach EUR 35 million or 7% of worldwide annual turnover.

Likely high risk: AI in a regulated product

The AI is a safety component of a product covered by Annex I. Full obligations for providers and deployers apply from 2 August 2028.

Likely high risk: Annex III use

The use falls in an area listed in Annex III. Full obligations apply from 2 December 2027. If you use a vendor’s system you are a deployer: follow the instructions for use, assign human oversight, keep the logs for at least six months and inform the people affected.

Likely not high-risk, under Article 6(3)

The use sits in an Annex III area but only performs a limited task. The provider documents this assessment before the system is placed on the market, registers it and gives the documentation to the authorities on request.

Transparency obligations apply

Under Article 50, since 2 August 2026: people must know they are dealing with AI unless it is obvious, AI-generated content is marked in a machine-readable way, and deepfakes are disclosed.

Likely minimal risk

No specific obligations beyond AI literacy. Most uses, such as drafting, summarising and analysis, fall here.

What to do next

Record the result in your AI register, with the purpose of the use and its owner. Whatever the level, Article 4 asks you to support the AI literacy of the people who use the system.

See the EU AI Act checklist for companies.

The four risk levels

Every use falls into one of four levels. Classify each use, not each product: the same assistant can be minimal risk for drafting emails and high-risk when it is used to screen job applicants.

LevelWhat it coversWhat it means for you
ProhibitedPractices banned by Article 5, such as social scoring or emotion recognition at work, and from 2 December 2026 AI that creates non-consensual intimate imagesStop using it
High riskUses listed in Annex III, such as recruitment, credit scoring and access to essential services, and AI in products covered by Annex IFull obligations from 2 December 2027 or 2 August 2028
TransparencySystems that interact with people, generate content or create deepfakesInform people and mark content, under Article 50, since 2 August 2026
MinimalMost other uses, such as drafting, summarising and analysisNo specific obligations beyond AI literacy

What the check does not cover

  • Your role: the obligations differ for providers and deployers. Most companies are deployers.
  • General-purpose AI models: their providers have separate obligations.
  • Systems already in use: high-risk systems already on the market before their date are covered only if they are significantly changed afterwards; those used by public authorities must comply by 2 August 2030.
  • Other law: the GDPR, sector rules and national law apply alongside the AI Act.

For a classification you can rely on, have each use reviewed with the Commission’s guidelines and your legal counsel. See EU AI Act compliance.

Sources

FAQ

Questions about AI Act risk levels

Is an assistant such as Claude, ChatGPT or Copilot high-risk under the EU AI Act?

It depends on the use, not on the product. Drafting, summarising and analysis are minimal risk. The same assistant used to screen job applicants falls in an area listed in Annex III and is likely high-risk.

Who has to classify an AI system?

The provider classifies the system it places on the market. A deployer needs to know the level of each use to meet its own obligations, and becomes a provider if it changes the purpose of a system so that it becomes high-risk.

When do the high-risk obligations apply?

From 2 December 2027 for the uses listed in Annex III and from 2 August 2028 for AI in products covered by Annex I, under the AI Act as amended by Regulation (EU) 2026/1744.

Is the result of this check legally binding?

No. It is a first indication for planning. The classification depends on the facts of each use and on the guidelines of the European Commission.

Classify every AI use. Start with a 30-minute briefing.

Tell us which AI systems you use today. We will reply with a clear next step.