How we use AI
A plain account of what artificial intelligence does on the Centrum Polonijne platform, where its limits are and what happens to your data.
Facts checked:
The technology behind the platform
- Claude by Anthropic is the language model behind the AI tutors and the task checks: conversation, corrections, and grading of written answers and photo tasks.
- Amazon Polly and ElevenLabs provide the synthetic voices: Polly voices words and dialogue lines, and the listening dialogues and texts are voiced role by role with ElevenLabs or Polly voices. They are not recordings of real people.
- Kasia, Emi, Adam and Mateusz are virtual characters, not people. Group and one-to-one lessons are taught by real teachers.
AI grades are approximate
- The AI’s scores and comments show you what to work on. They are not official exam results and do not predict your result in the exam.
- The AI can be wrong: it may miss a mistake or “correct” something that was right. If in doubt, ask a teacher.
- We wrote the tasks and the grading criteria from the publicly available exam format. Only the examination board issues official results.
What happens to your data
- To answer you or check a task, your text is sent to the Claude model — via Amazon Bedrock in the EU (Frankfurt) or via the Anthropic API (USA, under the EU Standard Contractual Clauses). Anthropic does not use this data to train its models.
- Voice audio is generated by Amazon Polly in the EU (Frankfurt). We prepare the listening recordings in advance from our own texts, so none of your data goes to ElevenLabs.
- On the platform, AI chat history is kept for 90 days after your last activity. Details are in the platform’s privacy policy.
- The “Try it — no sign-up” demo on this site asks for no name or e-mail: only the text you type is sent for checking, so please keep personal details out of it. Cloudflare Turnstile protects the demo from bots.