Translated from the Dutch article. The original publication date is retained.

Which EU AI Act obligations apply depends on the system, intended purpose, use and your role. The rules apply in phases. Check the current requirements before introducing a system or changing its use.

Start with purpose and use

The AI Act prohibits certain uses and sets additional requirements for high-risk systems. Transparency obligations and other rules also exist. These obligations can apply together. The size of your business does not determine the classification.

When is it high risk?

AI for recruitment, credit assessment or other decisions about people may fall within a high-risk category. The specific use, legal category and possible exceptions determine this. Assess it before building and obtain qualified advice where uncertain.

Transparency for a chatbot

People must understand that they are communicating with AI, at the latest at the first interaction, unless it is obvious. Additional disclosure rules apply to certain generated images and other content. A notice does not remove other obligations.

Internal tasks and AI literacy

For quotes, summaries, invoices and schedules, assess the purpose, data and consequences too. Internal use is not an automatic exemption. AI literacy measures must fit the roles, systems used and risks. A short training session is not a general guarantee. The GDPR still applies when personal data is processed.

How to work it out yourself

  • Record what each system does, which data it uses and who is responsible.
  • Assess whether prohibited-use rules, high-risk rules or transparency obligations apply.
  • Choose appropriate measures, document your reasoning and review it when things change.

Keep the reasoning behind your choices, agreements, checks and training. The documentation needed depends on the use and your legal role. One line or a certificate is not a general guarantee that all requirements are met.

Does a customer service chatbot fall under the EU AI Act?

Transparency obligations may apply. Make clear in time that people are communicating with AI, and also assess the use, data, other AI Act rules and GDPR.

When is an AI system high risk?

That depends on the legal category, intended purpose, specific use and possible exceptions. Recruitment and credit assessment can be examples. Seek qualified assessment when uncertain.

What should a business document?

Record purpose, role, data, risks, measures and checks. Exact obligations depend on the system and your role. A short note or certificate does not guarantee compliance.

Sources

  1. AI literacy: questions and answers ↗ · European Commission
  2. Transparency obligations under Article 50 of the AI Act ↗ · European Commission
  3. Navigating the AI Act ↗ · European Commission