1. Purpose and scope
Helpful Games AI is the collective name for the AI features in Helpful Games, operated by Elevspel AB. Helpful Games AI gives students explanations when they answer incorrectly and helps teachers create exercises. No personal data is sent to the AI service.
This policy describes which AI features exist, what data they use and which safeguards apply. It is intended for schools, school authorities, teachers, students and parents, and supplements our privacy policy and our data processing agreement.
2. How AI is used
Helpful Games AI is used in three features. None of them marks, assesses or grades students.
AI explanations for students. When a student answers incorrectly, they can click the “Explain” button to get a short explanation. No explanation is shown unless the student chooses it. Whether an answer is right or wrong is decided by the game’s own rules, not by AI. The explanation is a support and may occasionally contain errors.
AI support for teachers. Teachers can let AI suggest questions and answers for new games within a subject area. The teacher reviews, edits and decides what is published. No AI-generated material is published automatically. The feature is only available to teacher accounts, and the number of AI requests per session is limited.
Curriculum mapping. Game creators can get help from AI to map a game to a specific curriculum. This feature is only available for public games.
3. AI provider and data flow
Helpful Games AI uses Google Gemini through the Gemini API on a paid account, as well as locally installed language models. Requests are made from Helpful Games’ servers, never directly from the student’s or teacher’s device.
| Sent to Gemini | Never sent | |
|---|---|---|
| AI explanations | The question, the game’s metadata (such as title, subject, target group and description), the correct answer and any incorrect answers selected by Helpful Games | The student’s own answers (including free-text answers), name, age, user ID, email, school, class, IP address, results history |
| AI support for teachers | Subject, level, the game’s metadata (such as title, subject, target group and description) and the instructions the teacher writes | Personal data about the teacher or the students |
Since no personal data is sent, Google is not a data processor for the school’s student data. Teachers are asked not to include students’ names or other personal data in their instructions to the AI support.
Google does not use the content of the requests to train or improve its models, according to Google’s terms for paid API services. Google retains requests for 30 days to detect abuse. The Gemini API does not guarantee that processing takes place within the EU. Since no personal data is sent, no transfer of personal data to third countries takes place.
Local models run on Helpful Games’ own servers, and no data leaves Helpful Games.
All requests to the AI service are logged by Helpful Games. Logs are kept for 18 months and then deleted, unless otherwise agreed, in accordance with our terms of membership.
4. Risk classification under the AI Act
In our assessment, Helpful Games AI is not a high-risk AI system under Annex III of the EU AI Act (Regulation (EU) 2024/1689).
Annex III covers AI in education that determines access or admission, evaluates learning outcomes, decides the level of education a student should receive, or monitors students during tests. Helpful Games AI does none of these:
- Marking and results are calculated by the game’s rules, not by AI.
- The AI explanation does not affect the student’s results, level or which questions the student receives.
- The AI support for teachers produces content suggestions that the teacher reviews.
Helpful Games AI meets the transparency requirements of Article 50, including the requirement for machine-readable marking. See section 5.
The assessment is reviewed before new AI features are introduced.
5. Transparency and human oversight
AI-generated content is always marked, and people make all decisions.
- AI explanations are shown with the text “Generated by [Helpful Games AI]”, with the addition “and edited by the game creator” if the teacher has changed it.
- Everything created with Helpful Games AI is marked per question: questions, explanations and curriculum mappings. A question is marked as AI-generated, or as AI-generated and edited if the game creator has changed it.
- The game creator sees the marking in the editor, and Helpful Games sees it when reviewing games.
- The game description shows how many of the questions were created with AI, for example “7 of 10 questions were created with the help of AI”.
- The marking is also machine-readable in the page code, in accordance with Article 50(2) of the AI Act.
- The marking shows that the content comes from AI, not that it has been reviewed.
- Images and other media files generated with AI in games made by Helpful Games are clearly marked as AI-generated. Teachers who create media with their own AI tools are asked to label them “Generated with AI”. It is currently not possible to generate images or media files in Helpful Games.
- Using the AI features is optional. An AI explanation is only shown when the student clicks “Explain”, and the teacher can hide explanations for individual students.
- Students and teachers can always disregard the AI’s suggestions.
- The teacher is responsible for what they publish, whether or not it was created with AI support.
6. School control, safety and quality
School control. The teacher can hide explanations for missed questions for individual students. The setting is made per student and applies to all explanations, including those created with AI. Otherwise, the AI features are only used when someone actively chooses to: the student clicks “Explain”, and the teacher chooses to generate questions.
Content safeguards. The instructions to Gemini limit the answers to the subject of the question and to language that suits the rest of the game’s content. Google’s safety filters for dangerous content, hate speech, sexual content and harassment are set to the strictest level, which blocks responses even at low probability.
Quality control. All games made public are reviewed by Helpful Games before publication, including when the questions were created with AI support. Games shared privately with the teacher’s own students are reviewed by the teacher.
Feedback. Below each AI explanation, the student is asked “Was this helpful?” and can answer Yes or No. The answer is stored together with the question and the explanation, without any link to the student. We review explanations that receive many No answers and correct or block those found to be incorrect.
7. User responsibility, changes and contact
AI explanations are not infallible. Students and teachers should read them critically and answer No below explanations that do not help.
We update this policy when the AI features, the provider or the data flow change.
Contact us if you have further questions about our use of AI. See also our privacy policy.
The revision date of this page: 2026-10-21
