Privacy & your data
We don't keep a profile on you.
Worldrift is an independent studio based in Poland, in the European Union. EU law, including the GDPR, applies to everything we do with your data — the rights to see, download and delete it on this page are yours under that law, not a favour.
When you play, the game sends your story turns to AI providers through one server we run. That server keeps very little about you:
- a token balance
- a log of what each AI call cost
- any cloud saves you chose to make
No name, no e-mail, no tracking — it is billing data, and you can see all of it here, download it, or delete it.
Records are stored in the EU (Cloudflare's European data jurisdiction). Amounts below are in the game's token currency, the same number you see in the menu. For the raw records exactly as stored, the download at the bottom of the page is a single JSON file. Questions: contact@worldrift.ai.
Sign in
Sign in with the same account you play with. This sign-in lasts one hour and can only view and delete — it cannot play, spend tokens, or be used in the game.
With Steam we receive only your Steam ID; with Discord only your Discord user id. Neither gives us your name or e-mail. Both start with a quick Cloudflare bot check.
Signed in as (sign-in expires )
Your balance
Your account on our server is a handful of facts: how many tokens you have left, when the account was created, when it was last used, and whether you have bought a token pack.
What you have spent
One line per AI call — what it was for and what it cost in tokens — plus the free trial credit, any token packs you bought, and any refunds. Newest first.
| When (UTC) | What | Tokens | Reference |
|---|
Nothing has been charged to this account.
The reference is a random id for the call (or the provider's job id). It is not your identity — quote it if you ever write to us about a charge.
Image and text jobs in progress
Some image and long text generations run in the background at OpenAI. We keep a note of each one until it has been billed, and remove it 7 days later. Only the job id, the model and the cost are kept — never the image or the text itself.
| Started (UTC) | What | Model | Billed | Tokens | Job id |
|---|
Nothing in progress — no jobs are waiting to be billed.
Cloud saves
Stories you chose to save to the cloud. Saves made in the browser demo are stored as readable text and can be opened below. Saves made by the full game are encrypted on your computer before they are sent — we hold the key for you so the game can read them back, but the server never opens them.
No cloud saves — we hold none of your stories.
Your reports
When you report something in the game, we keep which message or picture it was, the reason you picked and your note, if you wrote one — never the story text or the picture.
| When (UTC) | Reason | What | Scenario | Your note |
|---|
You have not reported anything.
Usage counts
Our server also keeps anonymous counters — one line per request, with the provider that handled it, what it was for, how many tokens went in and out, the number of images, the size of any voice audio, and whether it succeeded. No text, ever. These lines are filed under a scrambled version of your id, not the id itself, and they expire on their own.
| When (UTC) | Provider | For | Model | Tokens in | Tokens out | Cached in | Cache writes | Images | Voice audio | Ops | Result | Weight |
|---|
Technical details
Server logs
Show every kind of log line that can mention you
| When it is written | The line, with your id filled in |
|---|
Where your story goes
Each AI call is forwarded to the provider the game picked for that job, with our API key attached and your identity removed — providers see the story text, not who you are. The last column is what each provider's own published policy says about keeping and training on that text; where a policy does not say, the cell is marked [verify] rather than guessed. Cloudflare, which runs our server and stores everything on this page, is listed first.
| Provider | Used for | What it receives | Their retention / training (published policy) |
|---|---|---|---|
| Cloudflare | Runs our server and database (EU jurisdiction), the usage counters, server logs, and the sign-in bot check | Everything on this page; story text passes through the server in memory and is not stored | Account, spending and saves: until you delete them. Usage counters: three months. Server logs: 3 days (Free plan) / 7 days (Paid plan). |
| Anthropic (Claude) | Narration and most text passes | The story context for the turn: recent narration, world state, your inputs | API inputs and outputs are deleted from Anthropic's backend within 30 days (Anthropic privacy centre, "How long do you store personal data"). Training use: not stated on that page [verify]. |
| OpenAI | Text passes (Responses API, background mode), image generation, voice | Story context; image prompts and reference images; narration text for voice | Not used to train OpenAI models unless opted in; abuse-monitoring logs kept up to 30 days. Background responses with store:false are deleted after the ~10-minute polling window (OpenAI "Your data" docs). |
| Google — Gemini API | Text passes, image generation (Gemini image models) | Story context; image prompts and reference images | Paid tier: prompts and responses are not used to improve Google products; they are logged "for a limited period of time" solely for abuse detection (Gemini API terms). The exact period is not published [verify]. |
| Google — Vertex AI | Text passes, image generation, voice | Story context; image prompts; narration text for voice | Inputs and outputs are cached for up to 24 hours by default; prompt logging for abuse monitoring applies to some account types (Vertex AI data-governance docs). Training use: not stated on that page [verify]. |
| Fireworks AI | Text passes (open models) | Story context | Does not log or store prompts or generations for open models without opt-in; does not train on API inputs without opt-in (Fireworks privacy notice). |
| DeepSeek | World-state bookkeeping passes | World-state facts and recent narration | Personal data is processed and stored in the People's Republic of China; retained "as long as necessary"; the policy offers an opt-out from training use (DeepSeek privacy policy). No fixed period is published [verify]. |
| WaveSpeed AI | Image generation and editing, upscaling, background removal | Image prompts; scene and character images | Generated media expires within 7 days (WaveSpeed data-retention policy). Prompt/input retention and training use: not stated [verify]. |
| fal.ai | Background removal (BiRefNet) | Character images | Account data is deleted 30 days after closure; retention of API inputs/outputs and training use: not stated [verify]. |
| AiEngine via RapidAPI | Background removal | Character images | Policy not located for either the vendor or the RapidAPI gateway [verify]. |
| Cartesia | Voice (text-to-speech) | Narration text | Content may be used to train Cartesia's models unless opted out via their form; zero data retention is Enterprise-only (Cartesia privacy policy, ZDR docs). Retention period: not published [verify]. |
| ElevenLabs | Voice (text-to-speech) | Narration text | Request history is kept by default (the docs state a 2-year default for conversation data); zero-retention mode is Enterprise-only. TTS-specific period: [verify]. |
| Inworld | Voice (text-to-speech) | Narration text | Terms: generally-available models are not trained on non-public customer materials. Retention period: not published [verify]. |
| Steam (Valve) | Sign-in only | Nothing from the game; Steam tells us your Steam ID | Not a data recipient — Steam is the identity source. See Steam's own privacy policy. |
| Discord | Sign-in only (browser demo, desktop app, this page) | Nothing from the game; Discord tells us your user id | Not a data recipient — Discord is the identity source. See Discord's own privacy policy. |
A provider only ever receives the calls the game routes to it; the spending list above shows what each call was for, and the usage counts show which provider handled it. Players who use their own API keys ("bring your own key") call providers directly and never pass through our server at all.
Download or delete
Download everything
One JSON file with every record exactly as stored: your account and balance, every spending line, every cloud save with its content (encrypted saves stay encrypted; your key is included so you can decrypt them), the usage counts, the log formats and the retention rules. This is the full technical version of this page.
Desktop app
The Worldrift desktop app stays signed in on each computer until you sign it out here. It then asks for a new sign-in the next time it starts, or within 6 hours if it is running.
Delete my data
Removes your account, balance, cloud saves and key, and any jobs in progress. The confirmation explains the few things that are kept or cannot be removed.
For the technically minded: the raw database layout
These are the migration files that created every table on our server, served straight from the deployed server itself. There are no other tables.
Loading…