This policy explains what data we collect when you use Mosio, what we use it for, who helps us process it, and your rights over your personal data under Indonesia's Personal Data Protection Law (Law No. 27 of 2022, "PDP Law").

Who we are

Mosio is operated by PT Teman Strategi Indonesia ("we", "us"). This policy covers the mosio.me website, the app.mosio.me dashboard, the MCP server at mcp.mosio.me, the Mosio API, and the Mosio plugin and connector for AI assistants such as Claude and ChatGPT. For the personal data we process to run these services, we are the data controller.

Data we collect

  • Account data: name, email address, email verification status, and password. Passwords are stored as hashes, never in plain text.
  • Session and security data: IP address, browser type (user agent), and sign-in time, used to keep your account secure.
  • Workspace data: workspace name, members and their roles, brand kit (business name, colors, logo, contact), briefs, video compositions, and rendered videos.
  • Workspace media: photos, video footage and sound you upload in the dashboard (the Media page, the Photos & videos panel in Social media, or the video editor) or from a chat with your AI assistant (the upload card, which sends the file straight from your device to our storage, or a ChatGPT attachment, which we copy from OpenAI's file storage through the link ChatGPT gives us), with their file names. From each file we make a copy for video (resized and converted, a still poster for footage, and a sound waveform for the editor). The copy is made without the file's metadata, such as where it was taken and the device make; the original file stays stored as uploaded and is used for social media posts.
  • Playground references and history: in the dashboard's Playground you can give a reference video, a video whose look you want yours to follow: uploaded from your device, or through the link to a public TikTok, Instagram, YouTube, X or Facebook post that you paste. For a link, our server fetches the video from that post (with yt-dlp, without signing in to the platform) and keeps it in the workspace's media with its link, marked as a reference: your AI assistant studies it through inspect_media like any other file, but it is never placed in a video and cannot be posted. When you press Susun prompt, Cari hook or Susun storyboard, the goal you wrote, the story pattern, tone, pace and hook you chose, a summary of the chosen brand (business name, voice, products and prices, contact), the chosen recipe, and what was measured in the reference video (its length, number of scenes and the start of its speech, not the video file) are sent to Meta (the Muse Spark model in standard mode, which does not use the content to train models) or, when chosen or when Meta is unavailable, to Anthropic (the Claude API) to write a directed prompt, three hooks, a storyboard with its caption, hashtags and cover text, or to rewrite one beat. For Cari hook, the opening words (up to 120 characters) of at most 5 public posts in your workspace's Discover whose views are far above their account's usual are sent too, with their view counts and scores, so the hooks learn from their pattern; that provider processes them on our behalf. Every prompt you send from the Playground is kept in its history (goal, brand, look and the prompt text) with the videos your assistant made from it, until you delete it.
  • My recipes: when you make a recipe from a reference video and tick the consent, our analysis server (Akamai/Linode, Jakarta) downloads a copy of the video through a signed link valid for at most 7 days, measures its cuts, pace, colours, loudness, on-screen text and speech, and sends the video to Meta (Muse Spark, the Meta Model API) to be described, in a mode that lets Meta use the video to train its AI models, as you agreed. The measurements and that description, not the video file, then go to Meta (the Muse Spark model, contributor mode) to write the recipe, on our behalf. The copy and its analysis on the analysis server are deleted once the recipe is written, or when you delete the recipe or the workspace; the recipe (name, description, motion plan, art direction) and when you consented stay in the workspace until you delete them.
  • Voice provider keys: the Google Gemini or ElevenLabs key you save in the dashboard to make voice-overs, dialogue, music and sound effects. The key is stored encrypted for that workspace, opened on our servers only when it is used, and never shown again; only its last four characters are visible. What is made is kept as workspace media, with each word's timing when ElevenLabs sends it.
  • E2B render key: on the Free plan, videos are rendered on your own E2B account. The E2B API key you save in Settings is checked with E2B, stored encrypted for that workspace, opened on our servers only to start that workspace's render sandboxes, and never shown again; only its last four characters are visible. E2B bills the sandboxes to your account under its own terms and privacy policy.
  • App access: the AI assistants and apps you authorize, the permissions (scopes) you grant, access tokens, and API keys. API keys are stored as hashes; only their prefix is visible.
  • Usage and credits: render history, video duration, and remaining credits.
  • Payment data: order number, pack, amount, and payment status. We never receive or store your card, bank account, or e-wallet details. Payments are handled by our payment provider.
  • Referral data: your workspace's referral code, the workspace that referred you, the payments that earned a commission, and the balance history. A code's owner sees only how many workspaces it referred and its commissions, not their names or content. The code from a referral link is kept in browser storage for up to 60 days, until you sign in. Anyone who opens a webinar ticket link with your code sees your workspace owner's name as the one who invited them; you see the commission from that ticket, not the buyer's name or email.
  • Webinar ticket purchases: name, email address, the ticket chosen, the Claude plan you pick in the form, order number, amount, the discount and referral code used, and payment status. Webinar tickets on mosio.me can be bought without an account. The code from a referral link on the ticket page is kept in browser storage for up to 60 days. Once a ticket is paid, your masked first name (for example D***t), the ticket type, and when you bought it may be shown to other visitors of the webinar page for 14 days; your email and full name are never shown. The webinar link in the ticket email is our own link that leads to the webinar room; we record when it is opened. When you order, the ticket page also sends your random PostHog ID and visit ID (see Cookies and tracking), so the order and its payment (ticket type, amount, discount and status, without your name or email) are recorded in PostHog with that visit and where it came from.
  • Class pre-orders: a class pre-order on mosio.me/kelas/ is bought the same way and records the same data as a webinar ticket (including the masked first name on the class page for 14 days and the PostHog record). The recording and the class open to the account that uses the email given at order.
  • Communications: the content of emails or messages you send us.
  • Research and production plans: Topic research through MCP sends your question and context to Anthropic and its web search service; cited answers are returned to your assistant for review. Keep secrets and customer personal data out of research queries. briefs, research sources and findings supplied by your assistant, open questions, storyboards, recipe choices, narration scripts and sound cues are stored when you or your assistant save a plan. Workspace members and assistants you connect can read them until you delete the plan or workspace. Saving a plan does not approve rendering or publishing.
  • Referral payouts: we store the bank name, account holder and encrypted account number, amount, status, dates, transfer reference, notes and recording operator for review and reconciliation. Platform superadmins can view referrer and referred customer identities, commission sources and payout history. Referrers still see only referral counts and their commissions. Payout financial records are not automatically erased when a workspace is deleted; deletion requests are reviewed through support.

Data from AI assistants you connect

When you connect Mosio to ChatGPT, Claude, or another AI assistant over MCP, the assistant sends only the data needed for the action being run. For example: the brief, the text and prices in a storyboard, the chosen recipe and format, the video composition, and requests to check or download a video.

Mosio does not receive the rest of your conversation history. Your conversations in the AI assistant are governed by that assistant provider's privacy policy, such as Anthropic's or OpenAI's. You can revoke an assistant's access at any time from the Mosio dashboard or from the connector settings in your assistant.

Browser extension

Our browser extension for Instagram and TikTok works only after you sign in to your Mosio account. It reads post data from the pages you open yourself: the account's name and follower count, view, like, comment, share and save counts, posting time, duration, caption, and the post's link and cover image. That data is sent to your Mosio workspace (Discover), where outlier and engagement scores are calculated, and the results are shown back on the page; comment text and video files are not sent. A copy is also kept in your browser so a sorted view is still there after the page reloads. Videos and images you download come directly from Instagram's or TikTok's servers to your device. The "To Mosio" button sends the post's link to Mosio, and our servers then download its video for you to study in Media. The extension's access token is stored in the extension; revoking access from the dashboard or signing out of the extension stops all of this. The extension also sends usage data to PostHog through mosio.me: the names of the features you use and their counts (for example sorting 50 posts by views, the number of files downloaded, the export format), the extension version, a random ID per installation and your Mosio account ID (not your email); this usage data never includes the accounts you research, captions, post links or comment text, and you can turn it off in the extension's Settings ("Send anonymous usage data"). Extension data in your browser is deleted with "Delete data" in the extension's Settings or by removing the extension; data in Discover can be deleted from the dashboard.

How we use data

  • To run the service: signing in, creating and rendering videos, and storing and delivering the results.
  • To prepare the media files you upload for the editor, for your AI assistant, and for the videos you render. From footage and sound we also make a contact sheet (up to 24 numbered stills from the video) and note its scene cuts, silences and loudness, without AI models. The speech in that footage and sound is also transcribed automatically (the words and their times) by the whisper.cpp speech recognition model, which we run ourselves on our render servers (Google Cloud in Singapore or E2B sandboxes in the United States); the recording is not sent to any other transcription service and is not used to train models. When you or your assistant ask for it (make_cutout), the person in a stretch of a video (at most 6 seconds) is separated from the background by the MODNet matte model, which we also run ourselves on the same render servers and which makes no new imagery; the cutout is kept with that file. The AI assistant you connect sees each file's name, kind, length and size; when it inspects a file (inspect_media), that contact sheet, the scene and silence times, and the transcript are sent to the assistant and processed by its provider. Files are processed on our isolated render servers, nowhere else.
  • To make a voice-over, dialogue, music, a sound effect or a designed voice when you ask for one, in the dashboard or through your AI assistant in a chat: the script or dialogue lines, chosen voice, model, delivery direction, and the description of the music, effect or designed voice are sent to Google (Gemini API) or ElevenLabs with your key. That provider processes the data for your account and under its own terms and privacy policy, not on our instructions. For example, on the Gemini API's free tier Google uses the text sent to train its models, and ElevenLabs' terms let it use content to improve its services.
  • Voice cloning, pronunciation dictionaries, audio cleaning and dubbing, only when you ask for them in the dashboard and only with your workspace's own key: for a clone, the sample files (ElevenLabs) or the source clip and consent recording (Google Gemini) are sent to that provider and then discarded from our servers, not stored; a cloned voice stays in your own provider account until you delete it in the dashboard (deleting it also withdraws the consent). Pronunciation dictionaries (words and how they are read) are kept in your ElevenLabs account. For audio cleaning and dubbing, the sound or video file from Media is sent to ElevenLabs; the result is saved to Media like an upload, and a dubbing stays in your ElevenLabs account until you delete it. We only record usage events (voice_cloned, audio_isolated, dubbing_started, dubbing_saved): the provider, the number and size of files, the language and the length, never the audio, a voice name or a script.
  • To count credits, process payments, and issue receipts.
  • To send transactional email, such as email verification, password resets, and payment receipts.
  • For webinar ticket buyers: to send the ticket, the webinar link and materials, and the bonuses included with the ticket to the email given at purchase.
  • For class pre-orders: to send the recording link and news of the class opening to the email given at order, and to open the recording and the class to the account with that email.
  • To keep the service secure and prevent abuse, for example through rate limits and detecting suspicious access.
  • To help you when you contact us, and to meet legal obligations.

Our legal bases are performing our service agreement with you, legal obligations, our legitimate interest in keeping the service secure, and your consent where we ask for it.

What we don't do

  • We don't sell your personal data.
  • We don't use your briefs, brand kit, or videos for advertising, and we don't share them with other users.
  • We don't use your content to train AI models.
  • Mosio runs no AI models of its own to generate images, video, or audio. Videos are rendered from compositions written by you or your AI assistant. Voice-overs, music and sound effects are made only when you ask (in the dashboard or through your AI assistant), by Google or ElevenLabs with your own key. The Gemini voice samples you can play were made once by us on Google Cloud from one sample sentence, the same for everyone, with no user data.

Cookies and tracking

The mosio.me website and the docs.mosio.me documentation use PostHog for analytics: the pages you open, where your visit came from (such as UTM parameters and the referring site), the buttons you click, how far you scroll, code or text you copy from the documentation and what you search for there, approximate location from your IP address (country and city), device type, and screen recordings of visits so we can see which parts are confusing. Form fields are always masked in recordings, and recordings are deleted after 90 days. PostHog keeps a random ID in a cookie and browser storage so a later visit, including to app.mosio.me, is recognized as the same visitor. We use no advertising cookies. In the app.mosio.me dashboard, PostHog also records the pages and buttons you use (with the same screen recordings; form fields, API keys, voice provider keys and QRIS codes are masked) and account events: sign-up, sign-in, credit purchases and their status, videos created and rendered, media files uploaded (their kind and size, not their content), prompts, hooks and storyboards written and sent from the Playground (kind of look, story pattern, whether a brand was used, writing time, and the AI tokens and cost, not their content), recipes made from videos (where the reference came from, its length, how long making it took and the analysis cost), voice provider keys saved (the provider only) and voice-overs, music, sound effects and designed voices made (provider, model, kind and length, not their content), brands and API keys created, and the tools your AI assistant calls. This data is linked to your account ID, name and email so we can see what helps and what gets in the way. When something fails on the website, in the dashboard or on our servers, PostHog also receives an error report: the error message, the code and page where it happened, and your browser and device type, linked to your account if you are signed in. Email addresses, tokens and the values you entered are masked in error messages before they are sent. If your browser sends a Do Not Track signal, browser analytics does not run. The app.mosio.me dashboard also uses the session cookies needed to keep you signed in. Our network provider may set technical cookies for security, such as bot protection. Links on our site may carry UTM parameters that show which button you clicked.

Who processes data for us

We use the following providers. They process data only on our instructions and only to run Mosio:

  • Cloudflare: network, website hosting, storage for videos and the media you upload (R2), and the relay (Cloudflare Workers) our requests to Meta pass through.
  • Neon: database, located in Singapore.
  • Server (VPS) providers: run the application.
  • Google Cloud (Singapore): video rendering, and preparing and inspecting media files.
  • E2B (United States): isolated sandboxes without internet access where videos are rendered and media files are prepared, inspected, transcribed and cut out. One sandbox per task, deleted as soon as the task ends. On the Free plan, that sandbox runs on your own E2B account.
  • DOKU: payment processing (QRIS and bank virtual accounts).
  • Resend: transactional email delivery.
  • Anthropic (United States): writing prompts, hooks and storyboards in the Playground from the goal you write when Claude is used, and writing a My recipes recipe from the reference video's measurements and description (not the video file) when Meta's model is unavailable (the Claude API).
  • Exa (United States): web search for video research (research_video_topic): the research question your AI assistant sends, such as the video's topic, and nothing else from your workspace.
  • Meta (United States): writing prompts, hooks and storyboards in the Playground (Muse Spark, standard mode: the content is not used to train models); describing reference videos for My recipes and writing the recipe from that analysis (Muse Spark). Meta may use those videos to train its AI models; a video is sent only after you agree.
  • Akamai/Linode (Jakarta): the analysis server for My recipes' reference videos.
  • PostHog: product analytics, error reports and visit recordings on the mosio.me website and the dashboard, located in the United States.

If you save a Google Gemini or ElevenLabs key, voice-over scripts and the descriptions of music, effects or designed voices are sent to that provider for your account (see How we use data). Otherwise, we disclose data to others only when required by law or by an order from a competent authority.

The Motion gallery at mosio.me/en/motion/ plays sample animations that load the GSAP library from jsDelivr and fonts from Google Fonts. Your browser contacts both services directly, so they receive your IP address and browser type under their own privacy policies. Other pages on mosio.me use fonts from our own server.

Data location and transfers

Our database is in Singapore, and some data may be processed by providers outside Indonesia. For every transfer abroad, we make sure the recipient provides an equal or higher level of protection, as the PDP Law requires.

Security

All connections use HTTPS. Passwords and API keys are stored as hashes, voice provider keys and E2B render keys are encrypted, video download links expire quickly, and rendering runs in isolation. Internal access to data is limited to what is needed. No system is perfectly secure. If a personal data protection failure occurs, we will notify you and the relevant authority within 3 x 24 hours, as the PDP Law requires.

How long we keep data

  • Account and workspace data are kept while your account is active.
  • Inspections that an AI assistant asks for before showing you a preview (still frames and layout findings), and composition bundles for rendering on your own computer, are deleted after one day. Download links expire after one hour.
  • Media files are kept until you delete them in the dashboard (the original, its copy, its contact sheet, its transcript and its cutouts go together). A copy you have added to a video stays in that video until the video is deleted. A workspace can keep up to 2 GB of media.
  • Voice provider keys and E2B render keys are kept until you delete them or their workspace is deleted.
  • When you delete a video, a workspace, or your account, we delete the data, including from backups, within 30 days.
  • Payment records are kept for as long as tax and corporate record-keeping rules require.
  • Webinar buyers' names and emails are kept until the event's recording and bonuses have been delivered; after that only the payment record remains.
  • Class pre-order names and emails are kept for as long as the class access lasts, because that access is opened to the email.

Your rights

Under the PDP Law, you have the right to:

  • be informed about and access a copy of your personal data;
  • correct data that is wrong or incomplete;
  • have your personal data deleted;
  • suspend or restrict processing;
  • withdraw consent you have given;
  • receive and use your data in a commonly used format.

Send requests to halo@mosio.me from your account email. We respond within 3 x 24 hours and may ask you to verify that the request really comes from you.

Children

Mosio is meant for business owners and content creators aged 18 and over. Users under 18 may only use it with the consent of a parent or guardian.

Changes to this policy

We may update this policy. The effective date at the top of this page will change when we do. We will tell you about important changes by email or in the dashboard before they take effect.

Contact

PT Teman Strategi Indonesia
Email: halo@mosio.me

This English version is provided for convenience. If it differs from the Indonesian version, the Indonesian version prevails.