Classified examples · Everyday AI tools · Po polsku
Developers use an AI coding assistant
Minimal riskDeployer
Engineers use GitHub Copilot, Cursor or a similar assistant to write and review code.
How we classify it
- The company uses the tool under its authority: deployer.
- Code generation is not prohibited and not listed in Annex III: not high-risk.
- Generated code is not content shown to the public as media, so Article 50 marking does not concern this use.
- Result: minimal risk. The practical issues are confidentiality of source code, secrets in prompts and review of generated code.
What applies now
- AI literacy for developers: review generated code, no secrets or customer data in prompts
- Business plan with no training on your code, where available
- Register entry and owner
What would change the answer
- If the team builds an AI feature for customers, the company becomes the provider of that AI system: see 'We build a chatbot on a model API'
Evidence to keep
- Register entry
- Engineering AI rules
- Training register
In the AI Compliance Kit: 02AI Acceptable Use Policy · 03AI System Inventory · 06AI Literacy Programme · 07Vendor Due Diligence Questionnaire.
Dates and sources
- AI literacy duty, Article 4: 2 February 2025
Legal basis: AI Act art. 4.
Similar examples
- Staff use ChatGPT, Claude or Gemini for drafting and summaries Minimal risk
- Microsoft 365 Copilot across email, Teams and documents Minimal risk
- AI note-taker that records and summarises meetings Depends on the use