Classified examples · updated 2026-09-23 · Po polsku
Is it high-risk? 20 common AI uses, classified
The same tools come up in every company: chat assistants, Copilot, website chatbots, CV screening, AI images. Each example below shows the reasoning step by step, what applies today, and what would change the answer.
Classification depends on the facts of each use. Use these as worked examples next to your own readiness scan, not as legal advice.
Everyday AI tools
Staff use ChatGPT, Claude or Gemini for drafting and summaries
Result: minimal risk. The duties that apply are AI literacy (Article 4) and, if personal data goes in, GDPR.
Microsoft 365 Copilot across email, Teams and documents
Result: minimal risk under the AI Act; AI literacy and data protection work are what matter.
AI note-taker that records and summarises meetings
Result: minimal risk if limited to transcription and summaries; switch off any emotion or engagement scoring of employees.
Developers use an AI coding assistant
Result: minimal risk. The practical issues are confidentiality of source code, secrets in prompts and review of generated code.
Customers and content
A vendor's AI chatbot answers customers on our website
Result: transparency obligations (Article 50), in force since 2 August 2026.
An AI voice agent answers the phone
Result: transparency obligations since 2 August 2026; deployer, or provider if the company built the agent itself.
Marketing creates images with an AI image generator
Result: transparency obligations for realistic images; minimal risk for obviously illustrative graphics.
AI writes blog posts, newsletters and product descriptions
Result: usually minimal risk when a person edits and signs off; disclosure needed for unedited public-interest text.
AI avatar or synthetic voice in training and promo videos
Result: transparency obligations; consent is needed where a real person's face or voice is used.
HR and employees
An AI tool screens and ranks job applications
Result: high-risk. The deployer obligations apply from 2 December 2027 (Digital Omnibus); GDPR applies now.
A video interview tool scores candidates' emotions
Result: prohibited since 2 February 2025, with the highest fines (up to 35 million euro or 7% of worldwide turnover).
AI writes job ads, or decides who sees them
Result: two different uses. The drafting is minimal risk; the targeting is high-risk, with deployer duties from 2 December 2027.
AI monitors and scores employees' productivity
Result: high-risk, with deployer duties from 2 December 2027; emotion inference is banned now.
Credit, fraud and insurance
AI scores consumers' creditworthiness
Result: high-risk, with deployer duties and the impact assessment from 2 December 2027; GDPR and consumer credit law apply now.
AI flags suspicious payments and fraud
Result: not high-risk under the AI Act; GDPR and payment rules still apply.
AI prices life or health insurance for individuals
Result: high-risk from 2 December 2027. Motor or property insurance pricing is not on this list.
Education and biometrics
AI grades tests or monitors exams in a training company
Result: high-risk from 2 December 2027; emotion inference is banned now.
Face recognition to open the office door
Result: verification for access is not high-risk under the AI Act, but biometric data is special-category data under GDPR Article 9.
Building AI products
We build a customer chatbot on a model API
Result: provider with transparency obligations since 2 August 2026; marking for systems already on the market by 2 December 2026.
We fine-tune an open-weight model for our product
Result: provider of the AI system in every case; general-purpose model obligations only above the compute criterion.