gemini-2.5-flash
Google — Vertex AI
- What it handles
- Listening to what you say and speaking the in-game reply back to you.
- Processed by
- Google, on Vertex AI.
AI Transparency
How this game uses AI
Teach My Little Sister How to Drive talks with you in real time. To make that work, your voice is sent to external AI services while a conversation is running. This page lists which services we use, when each one is used, what is sent to them, and what each provider does with it.
Last updated: 27 August 2026
Choose your region
We use different providers depending on where the game runs. Pick the region that applies to you.
You are reading the global version. In the Chinese mainland the game uses Alibaba Cloud and BytePlus services instead, under different terms.
Services we use
Two providers, three models. Google normally answers you. The OpenAI models take over when the first choice is too slow or cannot be reached, so the conversation does not break in the middle of a drive.
ChatGPT Realtime Mini 2.1
OpenAI — Realtime API
- What it handles
- The same conversation, taken over when the primary service is slow or unreachable.
- Processed by
- OpenAI, through the Realtime API.
ChatGPT Realtime Mini
OpenAI — Realtime API
- What it handles
- The last step, used if the first fallback fails as well.
- Processed by
- OpenAI, through the Realtime API.
How the game connects
The choice is made inside the game. A selector built into the game watches the connection, moves to the next service when the current one is too slow or cannot be reached, and retries the higher-priority service at regular intervals.
All three connections on the right carry your voice directly from the game to the provider. The connection to our server carries only the token.
- 1Before a conversation starts, the game asks our server to confirm that you own the game. Our server returns a short-lived token and nothing else. Your voice never passes through it.
- 2The game then connects straight to the provider and streams your voice there. gemini-2.5-flash on Vertex AI is used by default.
- 3If that connection becomes too slow or cannot be reached, the game switches to ChatGPT Realtime Mini 2.1, and to ChatGPT Realtime Mini if that one fails as well.
- 4At regular intervals the game retries the higher-priority service and moves back to it once it is healthy. A single play session can therefore use more than one provider.
What is sent
Whichever service is active receives the same thing: your voice, plus the small amount of game state needed to answer sensibly.
Sent to the provider
- Your voice, streamed while a conversation is active.
- The current in-game situation, such as the driving scenario in progress.
- Recent turns of the same conversation, so replies stay coherent.
Your voice is never sent to our servers. It goes from the game straight to the provider, and our server is involved only long enough to confirm your purchase and hand back a token — which it keeps no record of.
What each provider does with it
Each provider sets its own terms for the data it receives. The summaries below are ours; the linked documents are authoritative.
| Provider | Used to train models | Retention | Official documents |
|---|---|---|---|
| Google Vertex AI gemini-2.5-flash |
No Google states it will not use your data to train or fine-tune its models without your permission, and that by default it does not use customer data to train its foundation models. |
Governed by Google Cloud's data governance terms for Vertex AI. See the linked document for the current terms. |
Vertex AI data governance Google Cloud Service Specific Terms Google Cloud Privacy Notice |
| OpenAI Realtime API ChatGPT Realtime Mini 2.1 ChatGPT Realtime Mini |
No Data sent to the OpenAI API is not used to train or improve OpenAI models unless the customer explicitly opts in. We do not opt in. |
For the Realtime endpoint, OpenAI keeps abuse-monitoring logs for up to 30 days unless a longer period is required by law, and stores no application state. |
How OpenAI handles your data OpenAI Privacy Policy OpenAI Enterprise Privacy |
Guardrails on what the AI can say
Replies are generated live, so they are not written and checked in advance. These are the limits we put on them.
- 1System prompts keep the conversation on the subject of the game — coaching your younger sister through a driving lesson — and steer it back when it drifts.
- 2Further prompts hold the in-game roles in place so the scene stays coherent and stays immersive.
- 3If the game ever says something inappropriate, tell us on our Discord server and we will look into it.
What we keep
Separately from the providers, here is what Easy Fox Games holds.
We do not keep
- Recordings of your voice. Your audio is not sent to us in the first place.
- Any record of the play token we issue.
We do keep
- The replies the AI produces, read through the logging each provider offers inside our own account. We use them to improve the writing and the pacing of the game.
- Gameplay figures such as time spent per stage and number of failures — only if you agree when the game starts, and only to balance the difficulty.
Questions
If you want to know what was processed during your session, or you would rather not have your voice sent to these providers, write to us and we will help.