Privacy Policy

• Your name, and your email address or mobile number.
• Your organization or workspace name.
• A password, if you chose to set one. Passwords are stored only as a salted hash by our identity service; we cannot read them.
• Your service tier selection and configuration preferences.
• Billing address and payment details for paid plans. Payment details are handled by our payment provider (Section 9); we do not store full card numbers.

• Sign-in timestamps and session duration.
• Pages viewed and features used.
• Browser type and version.
• IP address.

We use this to keep the service secure and to decide what to improve. We do not use it to build advertising profiles, and we do not share it with advertising networks.

If you give us a mobile number, we use it only to send you the one-time codes you ask for and to let you sign in with it. We send one text message per code you request. Message and data rates may apply. Reply STOP to stop the messages or HELP for help, or write to support@arivaran.ai.

We do not sell mobile numbers. We do not share mobile numbers, or text-message opt-in data and consent, with third parties or affiliates for marketing or promotional purposes. Our SMS provider, Amazon Web Services (Section 9), receives a number only to deliver the code.

For Google sign-in we request exactly three OAuth scopes, all of which Google classifies as basic rather than sensitive or restricted:

Apple, Microsoft and GitHub sign-in request the equivalent basic identity scopes for those providers. If you use Apple's Hide My Email, we receive and store only the relay address Apple gives us, and we can operate your account with it.

• The provider name and the account identifier it issued, so we can match you to your Arivaran account on your next sign-in.
• Your email address, used as your account identity and for service notices.
• Your display name, shown to other members of your workspace.
• Your profile picture reference, if the provider supplied one.

We do not retain the provider's access or refresh token after the sign-in completes, and we do not use your sign-in to call that provider's other interfaces on your behalf. Connecting Microsoft 365 backup is a separate, explicitly authorized step described in Section 5.3; signing in with a Microsoft account does not enable it.

This identity data is kept for as long as your account exists, and is removed with your account under Section 8.2. You can unlink a provider or switch to an Arivaran password at any time from your account settings, and you can delete the account entirely.

Nothing about your use of Arivaran is sent to the sign-in provider. The provider learns that you signed in to Arivaran, because it performed the authentication, and it learns nothing else from us. We do not write to your account with that provider, and we do not post anything anywhere on your behalf.

Arivaran captures the contents of protected endpoint volumes (Windows in production; macOS and Linux endpoint support is in progress), which may include documents, spreadsheets and presentations, email archives and messaging data, database files, system configuration and application data, and any other files stored on the backed-up volumes.

We do not inspect, read, classify or index the contents of backup data. All content is processed as opaque byte sequences during deduplication and storage.

To be able to restore, we do store metadata about it: file names and directory paths, file sizes and timestamps, volume identifiers and disk geometry, snapshot timestamps, and SHA-256 block hash references. File names and paths can themselves be revealing, which is why they are covered by the same access controls and retention rules as the content.

Where the Kubernetes agent is deployed, we process resource manifests (Deployments, Services, Secrets, Custom Resources) as JSON bundles, Persistent Volume Claim data captured through CSI volume snapshots and the same content-addressed block pipeline as endpoint backups, and a change journal of API-server watch events for point-in-time restore.

Where a customer connects Microsoft 365 backup, we process, through the Microsoft Graph API using customer-authorized OAuth application credentials: Exchange Online mailbox items including message subject, sender, recipients and body or attachment content; OneDrive and SharePoint files and file versions; and Teams channel messages, files and tabs. We do not use this content for any purpose other than backup, restore, and the compliance features (eDiscovery search, legal hold) that a customer administrator invokes.

Where a customer enrolls mobile devices, we process device identifiers (Apple UDID or enrollment ID, or Android device resource name), device model and operating-system version, Apple Push Notification service tokens and the “push magic” values used to wake a device for management commands, queued management command payloads, and device state such as enrolled, lost mode or disabled. This supports device inventory, policy push, and remote lock or wipe at a customer administrator's direction.

When an administrator or an invited support user opens a remote desktop, terminal or VNC session to a protected endpoint, the on-screen content, keyboard and mouse input, and clipboard data exchanged during that session pass through the relay path described in Section 11. We do not log or retain session screen or input content beyond what is needed to establish and relay the live connection, except where session recording has been separately enabled. Where it has, the recording is treated as backup data and is subject to the same retention controls as Section 8.

Where document redline and merge is used, the submitted document content is processed in memory by the merge worker for the duration of the request and is not persisted beyond that operation's working files.

Where a customer routes AI agents through Beam Agent Access using the managed inspection gateway on an Arivaran-managed deployment, the relay terminates the agent's connection in order to forward it. On that route only, by default, we check each request body for secrets and credentials: API keys, access tokens, JSON Web Tokens and private keys. We do not inspect those bodies for personal data, and we do not classify prompts.

• The check records only the type and count of credentials found, never their content, on the authorization-decision audit record for that request.
• Each credential found is redacted before any log or record of the request is kept. A body that could not be inspected in full is withheld from such a record.
• The request itself is forwarded to the customer's chosen provider unchanged.

A customer administrator can opt their organization out on the dashboard's Agent Access page. Provider-blind and customer-side key-broker routes are never inspected. On a self-hosted relay this check is off unless that relay's operator turns it on. The full terms are in section 7A of the Data Processing Agreement; ask compliance@arivaran.ai for a copy.

When you open a support chat we process the conversation content, any attached files, and the account context needed to answer you. Some support conversations are processed with the help of a third-party large language model (Anthropic, see Section 9) to triage or draft a reply.

Support-request content may also be relayed into our internal team chat, which is a Matrix server we host ourselves rather than a third-party service. We mention it here because it is another place your support conversation can end up, not because a separate company sees it.

On some service plans, identical data blocks that happen to exist across multiple customers are stored only once. When two organizations back up the same Windows system file or the same common application, that data is stored once rather than twice.

• No organization can see, access, or learn about another organization's data through this process.
• Matching is based purely on whether the byte content is identical, verified by cryptographic hash.
• Your organization can opt out of cross-organization deduplication at any time.

Retention of backup data is set by the organization that controls it, through its retention policy. When backup data is deleted, file-to-block catalog entries are removed immediately, block hash references are decremented within 30 days, and physical block data is deleted once no remaining references exist. A block shared through cross-organization deduplication persists only while at least one customer still references it.

Account information, including the sign-in identity data in Section 4.2, is retained for the duration of your contract with us plus 12 months for billing and legal-compliance purposes. If you delete your account, you can ask us to complete that erasure earlier under Section 10.

Dashboard analytics data is retained for 24 months and then deleted automatically.

A credential-guardrail decision (Section 5.7) has no retention period of its own. It is kept for exactly as long as the authorization-decision audit record it is attached to, and is deleted with it.