EU AI Act quick check

EU AI Act quick check: which obligations apply to your AI use?

The EU AI Act assigns AI applications to four risk levels (from minimal through limited and high risk to prohibited practices) and attaches graduated obligations. Since 2 August 2026, the transparency obligations apply, for example for chatbots and AI-generated content (Art. 50). Regulation (EU) 2026/1744 postponed the high-risk obligations to 2 December 2027 (Annex III) and 2 August 2028 (Annex I). This quick check gives you a first classification of your use case in a few minutes.

Goshawk with a wide view, standing for an overview of regulatory obligations. Image generated with AI.
In 2 minutes

Quick check: your indicative risk class

Answer five short questions about your AI use. You get an indicative classification into one of the four AI Act risk levels.

1Do you use AI for any of these practices? Social scoring by authorities, manipulative influence, untargeted facial scraping, or emotion recognition in the workplace or in education.
2Is the AI used in any of these sensitive areas? Recruitment/HR, creditworthiness/insurance, critical infrastructure, law enforcement/justice, migration/border control, access to essential services, education/exams, or biometric identification.
3Does the AI interact directly with people (e.g. a chatbot) or generate text, images or audio (including “deepfakes”)?
4Does the AI process personal data in doing so?
5Do you develop or distribute the AI yourself (as a provider), not just deploy it (as a deployer)?

This quick check is a non-binding orientation and does not replace legal advice. The wording of Regulation (EU) 2024/1689 prevails; classification in individual cases should be done with expert advice.

Risk levels

What the four risk classes mean

Level 1

Minimal risk

No specific AI Act obligations; voluntary codes of conduct are recommended.

Level 2

Limited risk

Transparency obligations: people must be able to tell they are interacting with AI or that content is AI-generated.

Level 3

High risk

Strictest obligations: risk management, technical documentation, human oversight and conformity assessment.

Level 4

Prohibited

Practices banned under the EU AI Act (Art. 5). Such use is not permitted.

Checklist

The key AI Act steps as a checklist

For your internal preparation: compact, tick-as-you-go and save as PDF. Freely accessible, no sign-up.

As of October 2026 · Basis: Regulation (EU) 2024/1689 (EU AI Act), amended by Regulation (EU) 2026/1744

Prefer the checklist by email? That consent-based option is coming soon.

1 · Inventory your AI use cases

2 · Determine the risk class

3 · Obligations for high-risk systems

4 · Transparency obligations (limited risk)

5 · General-purpose AI (GPAI)

6 · Data protection & governance

Key dates

  • 02/2025: prohibited practices (Art. 5) and AI literacy measures (Art. 4).
  • 08/2025: obligations for general-purpose AI apply.
  • 08/2026: transparency obligations (Art. 50), penalties and supervision. New prohibitions (Art. 5) follow from 12/2026.
  • 12/2027 and 08/2028: obligations for high-risk AI under Annex III and Annex I (postponed by Regulation (EU) 2026/1744).
Guide: what you must label since August 2026

This checklist is a first orientation and does not constitute legal advice. The applicable wording of Regulation (EU) 2024/1689 prevails; classification in individual cases should be done with expert advice.

How Habicht helps

Risk class and evidence: built in instead of pieced together afterwards.

Habicht automatically classifies every AI use case by risk class and produces the corresponding documentation. The obligation becomes a button press.

FAQ

Frequently asked questions on the EU AI Act

In stages: prohibited practices and AI literacy measures since February 2025, obligations for general-purpose AI since August 2025, transparency obligations (Art. 50) since August 2026. Under Regulation (EU) 2026/1744, the high-risk obligations apply from December 2027 and August 2028 respectively.
AI in sensitive areas: such as recruitment, credit scoring, critical infrastructure, law enforcement or biometric identification. It is subject to the strictest obligations: documentation, oversight and conformity assessment.
No. It is a non-binding orientation and does not replace legal advice. Classification in individual cases should be done with expert advice.
Not necessarily. But a platform with built-in risk classification and automatic documentation reduces the effort considerably.
To providers who develop AI systems or place them on the EU market, and to deployers who use AI systems in a professional context. This includes companies that use an AI chatbot or AI assistant. Purely private use is exempt. Which obligations apply depends on the role and the risk level.
Since 2 August 2026 (Art. 50): chatbots and other AI systems that interact directly with people must be recognisable as AI. Deepfakes, i.e. realistic AI-generated images, videos or audio, must be disclosed, as must AI-generated texts that inform the public on matters of public interest and have not undergone editorial review. Providers of generative AI must additionally mark content in a machine-readable way.
Since February 2025, Art. 4 requires measures to build AI literacy among staff working with AI. Since the amendment by Regulation (EU) 2026/1744, companies must take such measures but need not ensure a specific level of knowledge for every individual. There is no certification requirement.
Up to EUR 35 million or 7% of worldwide annual turnover for prohibited practices, up to EUR 15 million or 3% for infringements of other obligations such as the transparency obligations (Art. 99). For SMEs and start-ups, the lower of the two amounts applies.

See Habicht in your own environment.

A short demo, tailored to your use case: on-premise, hybrid or in the EU cloud.