1. Who we are and what this policy covers
Rizzma Inc. operates the official HivemindOS website and hosted services described in this Privacy Policy. This policy covers the official HivemindOS application, hivemindos.app, official HIVE identity and community services, and other HivemindOS-managed services that refer to it.
HivemindOS is open source and can be self-hosted or modified. A third-party fork, deployment, model, wallet, runtime, integration, plugin, skill, or website is controlled by its own operator and is not covered by this policy. A managed product may provide an additional notice for practices specific to that product.
2. The local-first privacy model
The private local core is the default architecture, not a marketing wrapper around a cloud account. The dashboard server, agent bridge, durable dashboard state, chats, work records, schedules, local wallet records, shared-brain vault, search indexes, operational receipts, and local telemetry are designed to live on the device, workspace, or private fleet you control.
The app does not require a HivemindOS cloud account for local, bring-your-own-key, self-hosted, or local-model operation. A local-only workflow can remain entirely on the machines and private network you control when you select local runtimes and do not enable remote integrations.
The first-run desktop flow does not read private filesystem locations until you approve setup and local filesystem access. The component historically named the telemetry collector is a local agent bridge; its ordinary fleet and runtime telemetry stays on your device or private fleet rather than being uploaded to HivemindOS.
3. Information stored locally
Depending on the features you use, HivemindOS may store app preferences, terms acceptance, agent profiles, chat messages, attachments or file references, working directories, task and schedule records, notifications, runtime usage, budgets, approvals, wallet metadata, encrypted local wallet secrets, integration credentials, logs, evaluations, and operational receipts under the local HivemindOS application data directory, configured workspace, runtime directories, or shared vault.
Shared Brain content, memories, skills, indexes, retrieval records, and conversation mirrors remain in the configured local vault unless you enable a sync or remote-agent feature. Shared credentials are stored in the local HivemindOS environment file and may be copied to your own private fleet only when you configure that replication.
Production builds do not enable the app's detailed local developer telemetry by default. When local telemetry is enabled for development or diagnostics, it is written to a rotating file under the local application data directory. It is not HivemindOS cloud analytics.
4. Information that is not sent merely by opening the app
Installing or opening the app does not by itself send your prompts, responses, chat history, files, vault content, credentials, wallet private keys, seed phrases, local telemetry log, or detailed machine inventory to Rizzma Inc.
HivemindOS does not sell this local content or use it to build advertising profiles. Local data can still be exposed by malware, an insecure device, a misconfigured network, a plugin or skill, a provider you connect, a command you authorize, or another person with access to your system.
5. Limited automatic technical connections
The official website is delivered through Cloudflare, which necessarily receives technical request information such as IP address, approximate network location, browser and device type, user agent, requested URL, request time, referral information, and security events needed to deliver and protect the site.
Packaged desktop builds check the signed GitHub release endpoint when the dashboard loads and approximately every six hours while it remains open so the app can show that an update is available. The update is downloaded and installed only after you choose the update action. Source-checkout builds may query the configured Git remote for the current revision when showing source and fleet update status.
A connected feature may retrieve public availability, pricing, safety, payment, or policy configuration from HivemindOS or a provider when you open or use that feature. These technical checks can reveal ordinary network metadata such as IP address and request time, but they are not designed to include your prompts, responses, files, vault content, credentials, or wallet secrets.
6. When you choose a model or agent runtime
Local model servers can process prompts and outputs on your own machine. When you select a remote model, hosted runtime, coding agent, image or video provider, search provider, or paid API, the information required for that request is sent to the selected provider. This may include prompts, conversation context, system instructions, files or excerpts you attach, tool inputs, generated outputs, model and usage metadata, and technical request information.
If you choose an official HivemindOS-managed agent or model gateway, Rizzma Inc. and its infrastructure providers may process the request content, workspace or agent identifiers, authorization or entitlement data, usage, costs, tool results, error details, and receipts needed to provide, secure, bill, and audit that service. Provider-specific supplemental notices may apply.
In HivemindOS Mobile, current cloud AI recipients are OpenAI, Anthropic, OpenRouter and the model providers it routes to, and HivemindOS Cloud. HivemindOS Cloud is operated by Rizzma Inc. on Cloudflare; its free Swarm Scout service uses Modal Labs for inference, and its credit-backed requests use OpenRouter and the selected model provider, currently OpenAI or Anthropic. Realtime calls may send live microphone audio, transcripts, instructions, conversation context, and tool results to OpenAI, or through your paired hub and LiveKit to the configured voice provider. Image prompts and relevant context go to your paired hub and its configured image provider.
Before a cloud AI service receives personal data from HivemindOS Mobile, the app presents a permission screen that identifies the recipient or recipients, the categories of information being shared, and the purpose. Those categories can include typed or spoken prompts, selected photos and files, conversation history, agent instructions, workspace excerpts, and tool inputs or results. The request is not sent unless you choose Allow Sharing.
If you use your own provider key, the provider receives data under its own terms and privacy policy. HivemindOS does not make a remote provider private merely because its controls appear inside a local app.
7. Integrations, sync, remote machines, and communications
GitHub, Google, Slack, Discord, Telegram, email, browser automation, cloud deployment, external MCP servers, and other integrations communicate with the provider only after you configure, connect, authorize, or invoke them. The provider may receive account identifiers, OAuth tokens or API credentials, prompts, messages, files, repository data, destinations, and action details needed for the operation.
Hivemind Sync, Syncthing, Taildrop, shared-env replication, remote agents, and Hivemind Link move selected data between machines you configure. Those paths are intended for your own private fleet, but Tailscale, Syncthing relays, network operators, hosting providers, or other infrastructure may process connectivity metadata under their own policies.
Messages, posts, emails, deployments, commits, pull requests, and other outward actions go to the destination you choose and may become visible to recipients or the public. Approval and preview controls do not make an external action private after it is sent.
8. Your own devices: chats, settings, files, and managing your computer
When you use HivemindOS on more than one of your own devices, they can keep in step: your chats, a few settings (the voice that reads replies, your voice chat choices, and the look of result cards), and files you send from one device to another. Before any of this leaves a device, it is sealed with your account's brain key. Only your own devices hold that key: a phone or browser gets it only after you approve it on your computer by matching the number shown on both screens.
Our relay stores and passes these on only as sealed data it cannot open. For each item it sees its size, when it was written, and which account it belongs to, but never chat titles, messages, file names, file contents, your settings, or which device a chat or file came from. A chat with any message you chose not to save, or to keep on this device only (the Incognito choices), and shared rooms are never added. On the website, deleting a chat also takes it off your other devices' list.
Files sent with HiveDrop wait, sealed, for up to 7 days, or 24 hours if the upload never finished, and are then deleted. A file sent to one device is deleted once that device saves it; a file sent to all your devices stays until it runs out or you remove it. Chats stay until they are removed, or until your account holds about 400, when the oldest are taken out first. A note that a chat was removed is kept for 90 days so your other devices learn of it.
A phone or browser can also ask your computer for access, so you can manage it from there. The request carries the device's name, what kind of device it is, and what it asks for, so your computer can show them. Nothing is allowed until you approve it on the computer by tapping the number the phone or browser shows, and then only what you chose, such as its AI models, pictures and video, chats, files, automations, copy trading, background apps, or wallets. The computer's answer, and every request after it, is sealed with a key only that computer and that device hold, so the relay passes on bytes it cannot read. The browser keeps that key in its own storage on that device. Chatting with a model on your computer runs on your computer and uses no credits. You can change or remove a device's access at any time in Settings > Devices on the computer.
9. Honey rewards and usage metadata
Honey rewards are disabled by default and must be enabled by you. When enabled, HivemindOS may submit workspace ID, agent ID, token count or delta, model label, timestamp, source, and event ID to the official Honey ledger so usage can be deduplicated and credited.
The ordinary Honey usage observer is designed not to send prompts, responses, files, wallet keys, or detailed machine information to the Honey ledger. If you instead route model requests through the optional HivemindOS reward or compute gateway, that gateway necessarily processes the model request content and provider response needed to perform inference and measure usage.
10. Wallet and blockchain information
When you choose to connect or use a wallet, HivemindOS may read the wallet address you approve, active chain ID, public token balance, public staking state, contract state, transaction status, and other public blockchain information through your wallet or a configured RPC provider. Wallet addresses and blockchain activity are public identifiers and may be linkable to other activity.
When you authorize staking, trading, payment, transfer, withdrawal, or another blockchain action, wallet software, RPC providers, blockchains, facilitators, marketplaces, counterparties, and block explorers may process the transaction and related public information. Ordinary external-wallet connection does not give HivemindOS your seed phrase or private key. If you create or import a secret into the encrypted local wallet vault, the secret is stored locally and is used only for operations you authorize under the configured governance controls.
Official payment, marketplace, managed-credit, and entitlement services may receive wallet address, network, asset, amount, transaction or settlement proof, product or resource, tenant or agent identifier, and fraud-prevention metadata needed to verify and provide paid access.
11. HIVE identity claims, X OAuth, and Telegram Login
When you choose the public identity-claim flow, the service processes your wallet address, active chain ID, verification nonce, signed proof message, wallet signature, X OAuth state and PKCE data, redirect and timing information, and the X profile fields authorized by the requested scope. Those fields may include X user ID, handle, display name, avatar, profile URL, bio, and public follower, following, and post counts.
The OAuth access token is used to request the authorized X profile during the callback and is not stored in the identity database. The claim screen uses sessionStorage, a session-only browser store, to carry the wallet address through the redirect and removes it after the flow returns.
If you finish a claim with Telegram instead of X, the service processes the wallet proof described above and the fields Telegram's login widget returns and signs: Telegram user ID, first and last name, username, profile photo URL, sign-in time, and Telegram's signature, which the service checks with the HivemindOS bot's credentials. The service does not receive your Telegram password, phone number, or messages. To keep a claimed picture current, the service may ask Telegram for your current public profile photo through that bot and serve it from the identity service.
A verified wallet and X or Telegram association, wallet address, X handle or Telegram username, display name, profile picture, selected public profile information, public metrics, and verification time may appear in the public HIVE identity or community feed. Do not submit a claim if you do not want that association to be public.
12. Business agents, Google Search Console, Google Ads, and store connections
When you set up business agents on hivemindos.app and connect a Google account, HivemindOS requests read and write access to Google Search Console (the webmasters scope) and to Google Ads (the adwords scope), plus your Google account email address. We use this access only to provide the features you set up: reading your sites' search performance, sitemaps and indexing status; submitting your sitemaps; reading your Google Ads accounts, campaigns, keywords, search terms, costs and conversions; and creating, changing, pausing or budgeting campaigns, ad groups, keywords and ads within the daily spending limit you set.
Your Google refresh token is encrypted at rest and never shown to you, to other users or to your agents; each request uses a short-lived access token. Search Console and Google Ads data is read when an agent or you ask for it and is not stored in bulk. What we keep is limited to the connection, the Search Console property and Google Ads account you chose, and a record of each change an agent made with the value it replaced, so you can review and undo it.
To carry out the work you asked for, the relevant Google data is sent to the language model that runs your agent (a model provider you choose or HivemindOS's default provider) as part of that agent's request. It is not sold, not used for advertising, not used to train or improve generalized artificial intelligence or machine learning models, and not read by people at HivemindOS except with your permission, for security, or when required by law.
HivemindOS's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
If you connect a Shopify or WordPress/WooCommerce store, HivemindOS keeps the store's access credential encrypted and uses it only to read and update the product and page titles, descriptions and search listings your SEO agent changes, each change logged with what it replaced. Customer messages sent to your site chat or support inbox are stored with your account so the assistant can reply and hand conversations to you.
You can disconnect Google, which also revokes our access at Google, and disconnect your store at any time from your business team settings; you can also remove HivemindOS's access from your Google Account's security settings. Disconnecting deletes the stored credential immediately; the change log and conversations stay with your account until you delete your business data or ask us to.
13. Sales agent leads: people we find for businesses
The sales agent's AI model reads these details to judge whether a lead fits and to write a first email, and a second AI model checks each drafted email before it is saved. We show the results only to the business that asked, inside its HivemindOS account. HivemindOS does not send these emails: the business decides whether to write to anyone, and sends from its own mailbox. A business can download its own leads.
We keep a work address or search result we found for up to 90 days, so it can be reused for another business without looking it up again, and then delete it. A business's leads stay in its account until the business deletes them or closes its account.
If a business contacted you, or you would rather none did, you can remove your details at hivemindos.app/privacy/remove. We delete your address from every business's leads and from our stored results, and we will not look it up or add it again. If you use a work address, you can also ask us to leave out everyone at your company's domain. For access, correction or any other request about this information, email privacy@hivemindos.app.
Where data protection law requires a legal basis, we rely on legitimate interests: helping businesses reach other businesses with relevant, one-to-one messages, balanced by the controls above. Businesses that send these emails are responsible for the email laws that apply to them, including stopping when someone asks.
14. Gmail connections
If you connect Gmail, in the HivemindOS phone app, on hivemindos.app or in the HivemindOS desktop app, your agents, such as Scout, can use it in the ways you allow: when you ask them in a chat, or as part of an automation you set up, such as a summary of your inbox each morning. Reading is on to start: an agent can find and read your email, such as what someone sent you or whether a message arrived. You can also let agents write email for you. In the phone app under Apps, Integrations, Gmail you choose, for every agent or for each agent separately, whether it may read, write drafts, send, or send without asking you first. Agents in the desktop app can only read. HivemindOS never reads, writes or sends your email on its own: only when you ask, or when an automation you set up does what you set it up to do. You can turn an automation off at any time.
Google's consent screen asks you to let HivemindOS view your email messages and settings, and manage drafts and send emails. Both are asked for when you connect, so that allowing an agent to write later needs no second screen; HivemindOS uses only what your settings allow. It never deletes your email or changes its labels, and it does not read your contacts.
The connection is held by Composio, a service provider HivemindOS uses to connect apps. Composio keeps your Google tokens encrypted; they are never stored on your devices or in HivemindOS storage. HivemindOS identifies you to Composio by an internal account number, not by your name or email address. Composio is set not to keep the content of what it reads: it records only that a read happened, which kind, when, and whether it worked.
When you ask about your email in a chat in the phone app or on hivemindos.app, the AI model you are chatting with sees only your question. A search for the relevant emails is written from your words, the matching emails (at most eight: the sender, subject, date and a limited part of each message) are read, and a private AI model running in an attested secure enclave, which keeps nothing, writes the answer from them. No other AI model receives your email there. The answer is shown to you and kept in your chat like any other reply, and later replies in that chat are written without it; to do this, HivemindOS keeps a one-way fingerprint of each such answer, never the answer itself, for up to a year.
In the desktop app, an agent reads your email itself, whether you asked it in a chat or it is running an automation you set up. It searches for and reads the messages its task needs, and an AI model receives them. For an agent on HivemindOS's own engine, that model is always one that does not learn from what it is sent: the agent's own model when that model runs on your computer, runs in an attested secure enclave, or comes from a provider whose terms say it does not train on what it is sent. Otherwise that turn is answered by a private model that keeps nothing: Scout, HivemindOS's own model, inside an attested private enclave, or, when that is busy, another model in an attested enclave or from a provider that stores nothing it is sent, or a model on your computer. Agents that run as their own programs on your computer, such as Claude Code or Codex, read it with that program's own model, under your own account with that program's provider and its settings, such as whether it may use what you send to improve its models. What the agent writes from them is kept with its other work in the desktop app, like any other result. HivemindOS does not keep a copy of what was read.
When an agent writes an email for you, it is written from what you said in the chat, not from your mailbox. It is saved as a draft in your Gmail and sent only when you press Send on it, or straight away if you chose to let that agent send without asking. For each such draft HivemindOS keeps only its identifier, which agent wrote it and when, so that it can be sent once.
Gmail data is used only to do what you ask or what you set an automation up to do. It is not sold, not used for advertising, not used to train or improve generalized artificial intelligence or machine learning models, and not read by people at HivemindOS except with your permission, for security, or when required by law. Your email is never read into a shared chat room or an Incognito chat.
HivemindOS's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
You can disconnect Gmail at any time in the phone app under Apps, Integrations, Gmail, in the desktop app under Integrations, or on hivemindos.app under More, Integrations, which deletes the connection at Composio, and you can remove access from your Google Account's security settings.
15. Google Calendar and Google Drive connections
If you connect Google Calendar or Google Drive, in the HivemindOS phone app, on hivemindos.app or in the HivemindOS desktop app, your agents, such as Scout, can read it when you ask them in a chat, or as part of an automation you set up, such as a look at tomorrow's schedule each evening: what is on your calendar, when you are free or when an event is; or where a file is and what a document, spreadsheet or slide deck says. Agents can only read. They cannot create, change, share, download or delete your events or files, and HivemindOS never reads them on its own: only when you ask, or when an automation you set up does what you set it up to do. You connect each one separately, under Integrations.
Google's consent screen asks you to let HivemindOS see and download any calendar you can access, or see and download your Google Drive files. These are the read-only permissions Google offers for these products; HivemindOS uses them only as described here. Until Google has finished reviewing HivemindOS, Google first shows a notice that it has not verified the app.
The connection is held by Composio, a service provider HivemindOS uses to connect apps. Composio keeps your Google tokens encrypted; they are never stored on your devices or in HivemindOS storage. HivemindOS identifies you to Composio by an internal account number, not by your name or email address. Composio is set not to keep the content of what it reads: it records only that a read happened, which kind, when, and whether it worked.
When you ask about your calendar or files in a chat in the phone app or on hivemindos.app, the AI model you are chatting with sees only your question. For your calendar, the days you mean are worked out from your words, and the events in that period are read (at most forty: each one's title, time, place, the names of the people invited and a limited part of its notes). For your files, a search is written from your words, at most eight matching files are found (each one's name, kind, when it last changed and its link), and a limited part of the text of at most two Google Docs, Sheets or Slides among them is read. Other files, such as PDFs, are named only: nothing is downloaded or copied out of your Drive. A private AI model running in an attested secure enclave, which keeps nothing, writes the answer from what was read. No other AI model receives your events or files there. The answer is shown to you and kept in your chat like any other reply, and later replies in that chat are written without it; to do this, HivemindOS keeps a one-way fingerprint of each such answer, never the answer itself, for up to a year.
In the desktop app, an agent reads your calendar or files itself, whether you asked it in a chat or it is running an automation you set up. It reads the events or file text its task needs, and an AI model receives them. For an agent on HivemindOS's own engine, that model is always one that does not learn from what it is sent: the agent's own model when that model runs on your computer, runs in an attested secure enclave, or comes from a provider whose terms say it does not train on what it is sent. Otherwise that turn is answered by a private model that keeps nothing: Scout, HivemindOS's own model, inside an attested private enclave, or, when that is busy, another model in an attested enclave or from a provider that stores nothing it is sent, or a model on your computer. Agents that run as their own programs on your computer, such as Claude Code or Codex, read it with that program's own model, under your own account with that program's provider and its settings, such as whether it may use what you send to improve its models. Nothing is downloaded or copied out of your Drive. What the agent writes from them is kept with its other work in the desktop app, like any other result. HivemindOS does not keep a copy of what was read.
Calendar and Drive data is used only to do what you ask or what you set an automation up to do. It is not sold, not used for advertising, not used to train or improve generalized artificial intelligence or machine learning models, and not read by people at HivemindOS except with your permission, for security, or when required by law. Your calendar and files are never read into a shared chat room or an Incognito chat.
HivemindOS's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
You can disconnect Google Calendar or Google Drive at any time in the phone app under Apps, Integrations, in the desktop app under Integrations, or on hivemindos.app under More, Integrations, which deletes that connection at Composio, and you can remove HivemindOS's access from your Google Account's security settings.
16. Other connected apps (Slack, Notion, GitHub and more)
Besides Google, you can connect other apps your agents can use: as of October 1, 2026, Slack, Microsoft Teams, Outlook, OneDrive, SharePoint, Dropbox, Box, Notion, Confluence, Airtable, Asana, Trello, ClickUp, Monday, Wrike, Todoist, Linear, Jira, Productboard, GitHub, Bitbucket, Figma, Miro, Canva, Sentry, Supabase, PagerDuty, Contentful, Discord, Reddit, LinkedIn, Facebook, WhatsApp, Zoom, Webex, Dialpad, Calendly, Cal.com, HubSpot, Salesforce, Microsoft Dynamics 365, Zoho CRM, Attio, Zendesk, Zoho Desk, Gorgias, Mailchimp, Stripe, Square, Gumroad, FreshBooks, Zoho Books and YNAB. You connect each one separately, under Integrations, in the HivemindOS phone app, on hivemindos.app or in the HivemindOS desktop app. Signing in uses the sign-in that Composio, the service provider HivemindOS uses to connect apps, registered with that app, so the app's consent screen may name Composio rather than HivemindOS. It asks for the permissions that app sets for Composio's sign-in.
Your agents, such as Scout, can read a connected app when you ask them in a chat or as part of an automation you set up: for example the latest messages in a channel, your open issues or tasks, or a page. An agent can make a change there, such as posting a message, creating an issue or task, or updating a record, only after you approve that exact change: in a chat it is shown on a card with everything it will send, and nothing runs until you press Approve, once. Tools that delete things, such as a repository, a channel or a record, are not available to agents at all. HivemindOS never reads or changes these apps on its own.
The connection is held by Composio. Composio keeps your tokens for the app encrypted; they are never stored on your devices or in HivemindOS storage. HivemindOS identifies you to Composio by an internal account number, not by your name or email address. HivemindOS's Composio project is set not to keep the content of what it reads, but for apps connected through Composio's own sign-in Composio does not promise this in writing, so Composio's own privacy policy also applies to what passes through it.
When you ask about one of these apps in a chat in the phone app or on hivemindos.app, the AI model you are chatting with sees only your words. Which app, which of its actions and what to send are worked out from your words alone. What the app returns is read in a limited amount and given only to a private AI model that keeps nothing, which writes the answer; no other AI model receives it there. The answer is shown to you and kept in your chat like any other reply, and later replies in that chat are written without it. A change it proposes is shown for your approval and is not made until you approve it.
In the desktop app, an agent reads a connected app itself, whether you asked it in a chat or it is running an automation you set up, and the AI model that agent uses receives what it reads, under the terms of that model's provider. Choose a model that does not learn from what it is sent, such as a model on your computer, if that matters to you. A change still waits for your approval every time.
Data from these apps is used only to do what you ask or what you set an automation up to do. It is not sold, not used for advertising, not used by HivemindOS to train or improve artificial intelligence or machine learning models, and not read by people at HivemindOS except with your permission, for security, or when required by law. Your connected apps are never read into a shared chat room or an Incognito chat.
You can disconnect any of these apps at any time in the phone app under Apps, Integrations, in the desktop app under Integrations, or on hivemindos.app under More, Integrations, which deletes that connection at Composio. You can also remove access from the app's own settings for connected or authorized apps.
17. Public community and reward information
HivemindOS may collect public X posts, reposts, profile information, post URLs and timestamps, public engagement metrics, and public wallet associations relating to HivemindOS, HIVE, or configured community accounts. Community systems may generate contribution scores, score reasons, nomination status, reviewer notes, payout intent, HIVE amount, public transaction hash, and payment status.
This information is used to operate public community feeds, verify identity, prevent duplicate or fraudulent claims, review contributions, administer rewards, and provide an auditable public record. Source posts and blockchain records remain controlled by X and public networks even if HivemindOS removes its own copy.
18. How information is used
Depending on your location and the feature, information is processed to provide the service you requested; honor your choices and consent; secure accounts, payments, systems, and public claims; prevent abuse and fraud; troubleshoot and improve reliability; maintain records and audit trails; enforce the Terms; and comply with legal obligations.
Where consent is the legal basis, you may withdraw it for future processing. Withdrawal does not make earlier lawful processing unlawful and cannot erase data controlled by a public blockchain, a recipient, a social network, or another independent service.
19. When information is public or disclosed
Information may be disclosed to the public when you choose a public identity, community, publishing, deployment, repository, marketplace, wallet, or blockchain feature; to infrastructure and service providers that help deliver the selected feature; to connected providers and destinations you authorize; to advisers, auditors, transaction counterparties, authorities, or successors where reasonably necessary and lawful; and to a self-hosted operator if you use an installation controlled by that operator.
A service provider acting on behalf of Rizzma Inc. is required to provide the same or equal protection of personal data described in this policy. A provider that you connect independently, including through your own API key, also processes information under its own terms and privacy policy.
We do not sell personal information and do not share personal information for cross-context behavioral advertising. Verified claims and local app content are not used to build advertising profiles.
21. HivemindOS Browser extension
The HivemindOS Browser extension provides one disclosed purpose: letting you bring browser context into a HivemindOS side panel and send it to an agent you choose. After you invoke the extension, it may read the active page URL, title, selected text, readable page text, description, language, headings, visible links and controls, and the titles and URLs of open tabs. The extension does not continuously monitor browsing activity, and its privacy guard omits configured sensitive destinations such as banking, wallet, password-manager, checkout, payment, health, and tax pages.
The extension stores its dashboard URL, dashboard unlock token, selected agent, context mode, agent mode, and current session identifier in local, unsynchronized Chrome extension storage. These settings are not sent to Rizzma Inc. merely because the extension is installed, the browser is open, or you browse a page. When you choose Send, the extension transmits your prompt and the context shown in the side panel only to the HivemindOS dashboard address you configured. That dashboard may then send the request to the local or remote model, agent runtime, or provider you selected under its configuration and applicable policy.
The extension uses temporary active-page and script-injection access only after you invoke it, tab metadata access for its visible compare-tabs feature, local storage for connection settings, the browser side-panel interface, built-in access to the local HivemindOS dashboard, and an optional exact host permission when you connect another dashboard address. It does not request cookies, browser history, debugger, native messaging, downloads, bookmarks, or remote-code execution permissions. HivemindOS does not sell extension data, use it for advertising, transfer it to data brokers, or allow humans to read it except as specifically requested for support or when required for security or law.
The use of information received from Google APIs will adhere to the Chrome Web Store User Data Policy, including the Limited Use requirements.
22. Retention and deletion
Local data remains under your control subject to the app's deletion tools, your filesystem and backup practices, configured retention limits, and copies held by runtimes or destinations you selected. Deleting a local record does not delete an independently stored provider copy, recipient copy, public post, blockchain record, synced backup, or third-party log.
HivemindOS-managed services retain information only as long as reasonably necessary for service delivery, entitlement and payment records, claim integrity, fraud prevention, community and reward administration, dispute resolution, security, accounting, and legal compliance. Retention depends on the record and whether it remains active, disputed, security-relevant, or legally required.
In HivemindOS Mobile, choose Not Now to stop a proposed cloud request. You can revoke all future AI sharing in Settings > Privacy & AI Data; the app will ask again before another cloud request. Revocation does not erase information already retained independently by a provider. Requests to access or delete HivemindOS-managed information may be sent to privacy@hivemindos.app.
Public X content and blockchain records may remain public indefinitely. Cloudflare, GitHub, model providers, integration providers, and other third parties retain information according to their own configurations and policies.
23. Security
HivemindOS uses safeguards appropriate to the feature, including local-only defaults, private-network collectors, least-privilege connection design, signed wallet proofs, short-lived nonces, OAuth state and PKCE, encrypted local secret storage where offered, authenticated local routes, approval controls, spending limits, and provider access controls.
No system is completely secure. HivemindOS cannot guarantee that information will never be accessed, altered, lost, or disclosed without authorization. Keep devices patched, use strong authentication, review connected tools, restrict permissions, maintain backups, and never send a seed phrase or private key to a support contact.
24. International processing and privacy rights
Rizzma Inc. and connected providers may process information in the United States and other countries whose data-protection rules differ from those where you live. Where legally required, an applicable transfer mechanism or other lawful basis is used. Public blockchain and social-network data can be accessed globally by design.
Depending on your location, you may have rights to know or access personal information, correct it, delete it, restrict or object to processing, receive a portable copy, withdraw consent, appeal a decision, and lodge a complaint with a data-protection authority. You will not be discriminated against for exercising applicable privacy rights.
Submit a request to privacy@hivemindos.app and identify the relevant service, wallet address, or X handle without sending private keys, seed phrases, or unnecessary sensitive information. Proportionate verification may be required. HivemindOS may be unable to remove public blockchain records, source social content, recipient copies, or independent copies made by others.
25. U.S. state disclosures
In the preceding 12 months, official services may have collected the categories described above: identifiers; internet, network, and electronic activity; user-provided content; public social-profile and community information; blockchain, transaction, and commercial information; approximate location inferred from network requests; and inferences such as community contribution scores. These categories are used and disclosed for the operational purposes listed in this policy.
Rizzma Inc. does not sell these categories and does not share them for cross-context behavioral advertising. Sensitive personal information is not used or disclosed to infer characteristics about people. If practices materially change, this notice and any legally required opt-out or limitation mechanism will be updated.
26. Children
The official site and Services are not directed to children under 13, and wallet, staking, identity-claim, marketplace, and financial features are intended only for people legally able to use them. Rizzma Inc. does not knowingly collect personal information from children under 13 through these services. Contact privacy@hivemindos.app if you believe a child submitted information.
27. Changes and contact
This policy may be updated when services, providers, data flows, or legal obligations change. The effective date and version identify the current notice. Material changes will be highlighted or presented again where required.
Privacy questions and rights requests may be sent to privacy@hivemindos.app. General legal questions may be sent to legal@hivemindos.app. Do not include wallet secrets, private keys, or unnecessary sensitive information.