Bitebox Privacy Policy
Contents
1. Who we are and what this covers#
This Privacy Policy explains how GAS Labs, Inc., a Delaware corporation ("we," "us") collects, uses, shares, and protects personal information when you use Bitebox, including bitebox.gg, the creator Canvas, the Player at bitebox.gg/play/…, our MCP server and agent integrations, and related emails (the "Services"). It covers Creators (people who build experiences), Viewers (people who watch or play experiences, as guests or signed in), and anyone who contacts us. Terms not defined here have the meanings in our Terms of Service.
2. Information we collect#
2.1 Information you give us#
- Sign-in information. Your email address when you sign in with a one-time email link, or the profile information Google shares when you sign in with Google (typically your name, email address, and profile picture). We do not use passwords.
- Access codes. Creators enter an access code to unlock the beta. We check the code but do not store it on your profile.
- Creator Content. Prompts, uploaded images, video, and audio, node graphs, experiences, titles, descriptions, and edits. Uploads may include personal information about you or others, such as faces or voices.
- Purchase information. When you buy Credits, Stripe collects your payment details. We receive your Stripe customer reference, the pack you bought, its price, and the payment status. We do not receive your full card number.
- What you type into experiences. Some experiences ask Viewers to type or choose things. What you enter is saved as part of your play progress.
- Reports and messages. Reports of experiences or creators, blocks, support requests, and emails you send us.
2.2 Information collected as you use Bitebox#
- Play progress ("trails"). When you play an experience, we record a viewer identifier, which experience you are playing, where you are in it, playback position, which play-through it is, the events and choices along the way, and a snapshot of the experience's state. If you play as a guest and later sign in, your guest progress moves to your account.
- How you reached a story and your device type. In experiences we measure, we record the kind of link or site that brought you to the story (for example a NEUKO link, Instagram, TikTok, a search engine, or unknown) and whether you used a phone, tablet or computer. We read these from your browser when the story opens and keep only those categories, not the web address you came from or your browser's details.
- Guest identity. If you play without signing in, we create an anonymous guest identity so the Player can remember your progress.
- Canvas activity. Generations you run, which model, the outcome, and the Credit cost. Our usage records store the model, outcome, and cost, not your prompt text.
- Agent activity. If you connect an AI agent, we record the agent's client and label, the permissions and projects you granted, any expiry, when it was last used, when it was revoked, and which changes it made. We store a secure hash of its access token, not the token itself.
- Technical information. Our hosting provider receives standard request information such as IP address, browser type, and the pages requested. We use IP addresses briefly for rate limiting and do not store them in our database.
- Error reports. If something breaks, our error-monitoring service receives a report about the error with personal details removed.
- Product usage analytics. We use PostHog to understand how people use Bitebox. In the Canvas we record actions such as adding, connecting or deleting nodes, running a generation, previewing, saving and publishing, plus page views and clicks on Canvas controls with on-screen text masked. Currently these actions are linked to a pseudonymous ID derived from your account, not your name or email. We do not currently send email addresses to PostHog. In the Player we record anonymous play events: starting a story, scenes reached, which choices were shown and picked, time to choose, replays, restarts, completion and playback errors. We do not create a profile for Viewers or link play events to a Viewer's account. We do not send prompt text, what you type, uploads or payment details to PostHog. PostHog uses your IP address to estimate an approximate location, then discards the IP address. It stores that location with events sent from your browser: country, region, city, postal code, approximate coordinates and time zone.
- Upcoming Creator identification. After we notify Creators and this Policy's effective date has arrived, we plan to turn on identified Creator analytics in PostHog to understand and support your use of the Canvas. Once turned on, Canvas actions will be linked to your account ID, with your email address on your PostHog person profile only, not in event data. Until then, the current pseudonymous ID and no-email practice above continues. Viewers will remain anonymous.
- Upcoming Canvas session recordings. After we notify Creators and this Policy's effective date has arrived, we plan to record some Canvas sessions in PostHog so we can see where the Canvas is confusing or broken. A recording shows the Canvas layout, pointer movements, clicks, scrolling and page changes. Everything you type, all on-screen text, and all images and video are masked, so the recording does not show your prompts, titles, uploads or generated media. We do not record network requests or browser console logs. A recording is linked to the same Creator analytics ID as your Canvas actions. We never record the Player, Viewers, or sign-in and payment pages. Recordings are off when you turn product analytics off in Settings or your browser sends Global Privacy Control.
2.3 Information from others#
- Google, if you sign in with Google.
- Stripe, about your payment's status and disputes.
- Model Providers, about whether a generation succeeded, failed, or was blocked.
2.4 What we do not collect#
We do not collect precise location, such as GPS location from your device. PostHog estimates an approximate location from your IP address, as described in 2.2. We do not use facial recognition or create voiceprints or other biometric identifiers from uploads. We do not use advertising trackers.
3. How we use information#
We use personal information to:
- Provide Bitebox: sign you in, run generations, store and play experiences, remember your progress, and handle Credits.
- Give Creators analytics: visits, how far Viewers get, where they stop, endings reached, and time spent. Creators see these figures without Viewers' names, emails, or account identifiers.
- Run the beta: manage access and communicate about it. Our team can see your account's activity, such as sign-ins, generations, Credits and publishes, to run the beta and support you. Access is limited to authorized staff.
- Communicate with you: sign-in emails and important service or policy notices. We do not currently send marketing email.
- Keep Bitebox safe: prevent fraud, abuse, and security problems, handle reports and blocks, enforce our Terms, and report child sexual abuse material to the National Center for Missing & Exploited Children.
- Fix and improve Bitebox: diagnose errors and measure how people use Bitebox through product analytics (see 2.2).
- Meet legal obligations: respond to lawful requests, keep financial records, and handle copyright notices.
4. AI processing#
- Generations. When you run a generation, we send your prompt and settings, and for some models a reference image or video, to the provider that runs the model you chose, usually fal and for some models, such as Google's Imagen 4 Ultra and Runway's Gen 4.5, Replicate. Many models offered through fal are made by other companies, such as Google (Veo, Nano Banana, Imagen, Lyria), ByteDance (Seedance, Seedream), Kuaishou (Kling), MiniMax, Alibaba (Wan), OpenAI (GPT Image), Black Forest Labs (Flux), xAI, Runway, and ElevenLabs, and fal passes your request to them. The Canvas shows the name of each model you can choose. The provider runs the model and returns the result, which we store in your workspace.
- Other AI features. Text generation, media review, editing, experience conditions, and publishing features send prompts, text, images, frames, short clips, or transcripts from your project to AI models through fal, including its routing service (OpenRouter), or through Vercel AI Gateway. These models currently include Google's Gemini models, Anthropic's Claude Sonnet, DeepSeek models, TypeSafe AI's Jev, OpenAI's Whisper speech recognition, and Bria's background removal. Some editing features send a short clip from one step of your project to Google's Gemini models through fal and OpenRouter. In experiences that generate or check content during play, what a Viewer types or uploads is sent to fal with the experience's prompt, and from fal to models such as MiniMax, Flux, and Gemini. The providers that receive this content are listed in the AI providers table in Section 5.
- Provider terms. Each AI provider processes what we send under its own terms and privacy policy, which are linked in the AI providers table in Section 5. We use each provider's standard settings. When a provider passes a request to another company, such as the company that made or runs the model, that company's terms also apply.
- Training and retention by providers. Under their terms, AI providers may keep requests and results for the periods their terms describe, and longer where needed for safety, abuse prevention, or legal reasons. They may use data, including in anonymized, aggregated, or de-identified form, to improve their services. Some providers, and some companies whose models they route requests to, may use requests and results to train or improve AI models. fal, which runs most of our models, agrees in its API terms not to use our content to create, train, or develop its own products or services, except for models it marks "Pending Enterprise Ready". Section 5 summarizes each provider's terms.
- Our training. We do not use your Inputs, Outputs, or experiences to train our own AI models. If we ever want to, we will ask for your separate, opt-in permission first.
- Platform-Generated Materials. We use AI to write summaries, overviews, and logos for experiences from their content.
- Safety filters. Model Providers apply automated safety filters that may block content. We do not make decisions with legal or similarly significant effects about you solely by automated means.
5. How we share information#
We do not sell personal information and do not share it for targeted advertising. We share it only:
- With service providers that process it for us under contracts (see the tables below).
- With Model Providers, to run generations you request.
- Publicly, when you publish. A published experience, its title, your creator name, and its public details can be seen by anyone with the link, and its pages may appear in search results. We do not show your email address publicly.
- With agents you authorize, within the permissions you grant.
- For legal and safety reasons: to comply with law, protect people's rights and safety, and report child sexual abuse material.
- In a business transfer, such as a merger or sale of assets, subject to this Policy.
- With your direction or consent.
Service providers#
| Provider | What they do | What they receive | Where | Their terms |
|---|---|---|---|---|
| Supabase | Sign-in, database, sign-in emails | Accounts, guest identities, projects, play progress, Credit balances, agent permissions | United States (us-west-2) | Privacy |
| Vercel | Hosting | Request logs, including IP address and browser information | United States and global edge network | Privacy |
| Amazon Web Services (S3 and CloudFront) | Media storage and delivery | Uploaded and generated media, delivery logs | United States (us-east-1) and global edge network | Privacy |
| Stripe | Payments | Payment details, customer reference, purchase details | United States | Privacy |
| Sentry | Error monitoring | Error reports with personal details removed | United States | Privacy |
| PostHog | Product analytics | Currently, usage events with a pseudonymous Creator ID or an anonymous Viewer ID, plus page and device information. Events sent from your browser also carry an approximate location that PostHog estimates from the IP address (see 2.2). After notice and the effective date, when Creator identification is turned on, Creator events will use the account ID and email will be on the Creator person profile only, not in events. After notice and the effective date, masked Canvas session recordings (see 2.2). No IP address, prompts, typed input, uploads or payment details | United States | Privacy |
AI providers#
These providers process what we send under their own terms, linked below. The retention and training column summarizes those terms; the linked terms govern and may change.
| Provider | What they do | What they receive | Where | Retention and training under their terms | Their terms |
|---|---|---|---|---|---|
| fal (Features & Labels, Inc.) | Running AI models, including models made by other companies (see Section 4), speech recognition (Whisper), object masking (SAM), and background removal (Bria), and carrying requests to OpenRouter | Prompts, settings, reference images, video and audio, frames, short clips, transcripts, Viewer input in experiences that generate or check content during play, results | United States and other countries, including the locations of companies whose models it runs | Keeps request inputs and results for 30 days by default, and stored media for the period set for it. Its API terms say it will not use our content to create, train, or develop its own products or services, except for models it marks "Pending Enterprise Ready", for which that promise and its security and data processing commitments do not apply. May use anonymized or aggregated usage data to improve its services and develop its models. Companies whose models it reaches through their own APIs receive requests under their own terms. | Terms, API terms, Privacy |
| Replicate | Running some AI models, currently Imagen 4 Ultra, Gen 4.5, p-video, and video object masking (SAM 3) | Prompts, settings, reference images and video, results | United States | Removes the inputs, results, files, and logs of predictions made through its API after one hour by default. May use aggregated, anonymized data to improve its services. Models published by others on Replicate are subject to their own terms. | Terms, Privacy |
| OpenRouter (through fal) | Routing requests to AI models for text generation, media review, and editing features | Prompts, text, images, frames, short clips | United States and the locations of the model providers it routes to | Does not store prompts or results by default but keeps request details such as token counts. The model providers it routes to keep requests for their own periods, and some may train on them, under their own terms. | Terms, Privacy |
| Google (through fal, Replicate, OpenRouter, and Vercel AI Gateway) | Gemini models (currently Gemini 2.5 Flash, 2.5 Flash Lite, 3.1 Pro, and 3.8 Flash) for text generation, media review, editing, play-time checks, and publishing features; Nano Banana, Imagen, Veo, and Lyria for generations and logo images | Prompts, text, images, frames, short clips, transcripts, Viewer input in experiences that check content during play | Google Cloud locations | For paid Gemini API use, Google says it does not use prompts or responses to improve its products and logs them for a limited period to prevent abuse. On Vertex AI, Google does not train on customer data without permission and may cache it in memory for up to 24 hours. Which terms apply depends on the provider that carries the request. | Gemini API terms, Cloud privacy |
| Anthropic (through fal and OpenRouter) | Claude Sonnet, for text generation | Prompts, text, images | Model provider locations | Does not train on content from its commercial services. Deletes API inputs and results within 30 days, except where it must keep them longer to enforce its usage policy, for up to 2 years when flagged, or by law. | Commercial terms, Privacy |
| Vercel AI Gateway | Routing requests to AI models for experience conditions and publishing features | Condition text, stills, transcripts, titles, and episode text | United States and the locations of the model providers it routes to | Does not keep prompts or results after a request completes and does not train on them. The model provider that runs each request handles it under its own terms. | AI Gateway, Privacy |
| Model hosts through Vercel AI Gateway | Running DeepSeek models (currently DeepSeek V4 Flash) that turn experience conditions into rules | Condition text | Locations of the host Vercel AI Gateway selects | Varies by host. Some hosts may keep requests or use them for training under their own terms. | Hosts by model |
| TypeSafe AI | Jev, which suggests genres for published experiences, through Vercel AI Gateway | Series and episode descriptions | United States | Does not train or fine-tune models on prompts. Keeps them only as long as needed to provide its service or as the law requires. | Terms, Privacy |
Discord receives information only if you choose to join our Discord server through the invite link.
6. Cookies and browser storage#
We use cookies and browser storage that Bitebox needs to work. PostHog analytics IDs are kept in memory only, not in cookies or local storage. In experiences we measure, the Player also keeps a first-party device return ID, described below. We do not use advertising cookies.
| Name | Purpose |
|---|---|
| patchwork_editor_session | Keeps Creators signed in to the Canvas |
| patchwork_viewer_session, patchwork_viewer_guest_proof | Keep Viewers signed in or connected to their guest identity |
| pw_pkce_* | Secures the Google sign-in handshake |
| bitebox_oauth_* | Links a distribution account you choose to connect |
| Supabase sign-in token (browser storage) | Keeps the Player signed in |
| patchwork.auth.display-email:* (browser storage) | Shows which email you used to sign in |
| patchwork.editor.* (session storage) | Remembers access-code and sign-in steps during your visit |
| Canvas PostHog analytics ID (memory only) | Currently keeps your pseudonymous Creator analytics ID for this visit. After notice and the effective date, when Creator identification is turned on, it will use your account ID; it stays in memory only |
| Player analytics ID (memory only) | Keeps the anonymous analytics ID for this visit only |
| Player device return ID (first-party local storage) | Measures returning players on this device in experiences we measure. It is sent only to our measurement collector, never to PostHog; never linked to an account; it rotates every 90 days |
Creators can turn product analytics, including Canvas session recordings, off in Settings. Viewer play events are sent only when measurement is on.
Do Not Track and Global Privacy Control. We honor Global Privacy Control by turning off product analytics in that browser.
7. How long we keep information#
| Information | How long |
|---|---|
| Creator accounts | Until you delete your account, except records we must keep for tax, security, copyright, or legal claims |
| Projects in trash | Until you delete them, or until we delete them automatically after at least 14 days in trash |
| Creator Content (uploads, outputs, experiences) | Until you delete it or your account, except copies kept for the reasons above |
| Play progress (trails) | Until you delete it or the account it belongs to is deleted |
| Guest identities | Until the guest session ends or you sign in and keep the history |
| Product analytics events | 12 months in PostHog, then deleted |
| Canvas session recordings, once turned on | 30 days in PostHog, then deleted |
| Aggregated analytics | May be kept; it cannot identify you |
| Agent permissions and activity records | While the permission is active, and afterward as needed for security and abuse review |
| Purchase and Credit records | For as long as tax, accounting, and dispute rules require |
| Reports, blocks, and enforcement records | As needed to investigate and to meet legal duties |
| Error reports | For the period the error-monitoring provider retains them |
| Hosting request logs | For a short operational window, then deleted |
| Media delivery logs | For the period the delivery provider retains them |
| Backups | Until that backup window expires |
| Material reported to authorities | As required by law |
8. Security#
We use safeguards suited to the information we hold, including encryption in transit, access controls in our database, private media storage served through signed delivery, scoped and revocable agent permissions with hashed tokens, and error reports with personal details removed. No system is perfectly secure. If a security incident affects your personal information, we will notify you and regulators as the law requires.
9. Your choices and rights#
Wherever you live, you can ask us to:
- Access or export your account information and Creator Content. Access and export requests also cover PostHog data for the Creator's current hashed ID and, once Creator identification is turned on, their account ID and any earlier hashed ID.
- Correct inaccurate information.
- Delete your Creator Content or your account. Account deletion removes your sign-in, projects, experiences, media, play progress, and agent permissions, except records we must keep for tax, security, copyright, or legal claims. Deletion requests also cover PostHog data for the Creator's current hashed ID and, once Creator identification is turned on, their account ID and any earlier hashed ID.
- Delete your play progress as a Viewer.
You can also revoke agent access at any time in Settings.
To make a request, email legal@neuko.ai from the email address on your account. Guest Viewers can include the identifier shown in the Player's settings so we can find their progress. You may use an authorized agent where the law allows. We will not treat you differently for exercising your rights. If we deny your request, you may appeal by replying to our decision, and if your appeal is denied, you may contact your state attorney general or data protection authority.
10. U.S. state privacy notice#
This section applies to residents of California and other states with comprehensive privacy laws, to the extent those laws apply to us.
| Category | Examples | Sources | Shared with |
|---|---|---|---|
| Identifiers | Email address, Google profile, account and guest identifiers | You, Google, your device | Service providers |
| Commercial information | Credit purchases | You, Stripe | Stripe, service providers |
| Internet activity | Play progress, Canvas and agent activity, product usage events | Your use of Bitebox | Service providers |
| Geolocation data (approximate, not precise) | Approximate location estimated from your IP address (see 2.2) | Your IP address, through PostHog | Service providers |
| Audio and visual information | Uploads and generated media | You | Service providers, Model Providers, the public if published |
| Sensitive information | Sign-in credentials | You, Google | Service providers, only for signing you in |
We do not sell personal information or share it for cross-context behavioral advertising, and have not done so in the past 12 months. We do not knowingly sell or share the information of anyone under 18. We use sensitive information only to provide the Services. Subject to applicable law, you may request to know, access, delete, correct, or port your information, and to opt out of sale, sharing, targeted advertising, and certain profiling. We do none of these. California residents may also ask about disclosures for direct marketing; we make none.
11. People in the EEA, UK, and Switzerland#
Legal bases. We process personal information to perform our contract with you (accounts, generations, Credits, playing experiences, agents you authorize); for our legitimate interests in security, abuse prevention, Creator analytics, measuring product usage, improving Bitebox, and promoting published experiences, which we balance against your rights; with your consent where the law requires it for analytics storage, and where otherwise required; and to meet legal obligations.
Your rights. You may access, correct, delete, restrict, or port your information, object to processing based on legitimate interests, and withdraw consent. You may complain to your local data protection authority (in the UK, the Information Commissioner's Office).
International transfers. We are based in the United States. When we transfer information from these regions, we rely on the European Commission's Standard Contractual Clauses (with the UK Addendum where relevant), the EU-U.S. Data Privacy Framework where the recipient is certified, or another lawful mechanism.
12. Children#
Bitebox is for adults 18 and older. We do not knowingly collect personal information from anyone under 18. If we learn we have, we will delete it. Parents or guardians can contact legal@neuko.ai.
13. Other services#
Bitebox links to and works with services such as Google, Discord, Instagram, TikTok, YouTube, Stripe, and AI agent providers. Their privacy practices are governed by their own policies.
14. Changes#
We will post updates here and change the effective date. For material changes, we will notify Creators by email or in the product before the changes take effect. Previous versions are available at /legal.
15. Contact#
GAS Labs, Inc., a Delaware corporation 240 S. State St., Westerville, OH 43081 Privacy: legal@neuko.ai
Questions? Contact legal@neuko.ai