Privacy Policy
What Aurora collects, where it is stored, and what you can ask us to do with it.
- Last updated
- October 29, 2026
- Version
- 2026-10-29-2
1. Data Controller
Aurora is developed and operated by Ogard Labs (organisation number 996 198 510, Norway) ("we", "us"). We are the data controller for personal data we collect about you.
Contact us:
Email: contact@aurora.ink
Address: Ogard Labs, Stavanger, Norway
Organisation number: 996 198 510
2. Data We Collect
- Account information: email address and display name you provide at registration.
- Content you create: pages, notes, tasks, and other objects you store in Aurora.
- Usage data: login timestamps and last-active time — used only for security and debugging.
- Anonymous usage statistics, if you opt in: your country (derived from your device timezone, not your IP address), a daily usage count. We never collect IP addresses, device or account identifiers, any other personal data as part of this. This is off by default until you answer the one-time prompt; always changeable in Settings → Profile.
- Local storage: we use
localStorage,sessionStorage, andIndexedDBexclusively for app functionality (preferences, offline cache, short-lived handoff state). We use no third-party tracking cookies. - Google account and Google Calendar data, only if you use them: if you sign in with Google, Aurora receives your Google email address, name, profile picture address and Google account identifier (the
openid,emailandprofilepermissions). They become part of your Aurora account and follow the account retention in section 4. If you connect Google Calendar in Settings → Calendars, Aurora asks for one more permission, read-only access to your calendars (https://www.googleapis.com/auth/calendar.readonly). It then reads your list of calendars and the events in them (title, time, place, description, attendees and status, as Google returns them), and the email address and name of the connected Google account. We use this only to show your events, read-only, in Aurora's calendar view and to name the connected account in Settings. Aurora never creates, changes or deletes anything in your Google Calendar. With this connection, the access and refresh tokens, the calendar list and the events are saved only in your browser's local database (IndexedDB) on your device. They are not stored in AuroraCloud and are not part of cloud sync or our backups. To fetch events, your device sends its token and the request through our servers in Norway, which forward it to Google and pass the answer back; with this connection our servers do not store the token or the events. If you use folder sync, the connected account's email address and calendar names (not tokens or events) are written to the folder you chose. To remove this data, use Remove on the account in Settings → Calendars, which deletes the tokens, calendars and events from your device; clearing your browser data does the same. Removing the account in Aurora does not by itself withdraw Aurora's access at Google: do that at myaccount.google.com/permissions. There is a second way to connect Google Calendar, in Settings → Preferences → Integrations, and it keeps the connection on our servers. There Aurora saves in our database the connection's access and refresh tokens (encrypted), the name or email address of the Google account, the read-only permission granted, and when the connection was made and last refreshed. Events are not saved: when a request is made through this connection, our servers fetch the data from Google and pass it on to you without keeping it. We record only that the request happened (its time, its path, which can contain a calendar's identifier, and Google's status code) in a security log kept for up to 90 days. The tokens are kept until you disconnect or delete your account. Only you can use the connection through Aurora; the administrators who run our servers can technically reach the database and the key that decrypts the tokens, and our staff do not read your Google data. Disconnect (Settings → Preferences → Integrations) asks Google to revoke the permission and deletes the connection and its tokens from our database; copies in our encrypted backups are deleted within 30 days. Deleting your Aurora account also deletes the connection from our database, but it does not ask Google to revoke the permission, so disconnect first or revoke it at Google. In both cases we do not share Google data with anyone, do not use it for advertising, and do not send it to AuroraAI or any other AI provider. To remove the Google sign-in data held in your account, delete your account in Settings → Account. - Hosted import column review: if you turn on the optional AI review when importing, that import sends only the columns whose type the importer was unsure of — never one it could read as a definite type, such as a date, a number, a checkbox, a web address, an email address, a phone number, an attachment or a link to another database — and for each of those: the column's name, the type the importer guessed for it, up to 12 of the distinct values in it, each cut to the first 80 characters with a mark showing it was cut, how many rows hold each of those values, and how many rows the database has and how many of them are filled in. Each column also carries an opaque marker for the database it came from, so columns of one database can be told from another's; the marker stands in for the database's name and file path, which are not sent. Because this review runs on the same hosted route as AuroraAI chat, the request also carries your workspace's identifier as the provider's caller field, together with the model and routing details the server selects. This goes through the same Requesty EU gateway and the same
policy/auroradocspolicy as built-in AuroraAI chat. At most 120 such columns are sent for one import. Page text is never sent. The option is off by default and is asked for each import; if you decline it, if no provider is available, or if the published policy does not yet cover this review, the import uses its own column-type heuristics and nothing is sent. If you choose your own provider key instead, your device sends that request directly to that provider. - Hosted vault property review: this is a separate step from the column review above, with a policy gate of its own, but one opt-in covers both. The import screen shows a single option before each import, off by default, that names the database-column case and the vault-property case together and applies to whichever the files you pick turn out to be. An Obsidian vault declares no property types at all, so if you turn this review on, that import sends only the properties it could not settle to a definite type: the ones your notes contradict each other about, and the ones every note holding them writes as ordinary text. It never sends one that the notes holding it all write the same way as a date, a number, a checkbox, a web address or a list, so how many of your notes use a property decides nothing on its own: of two properties only a handful of your notes use, one can be left out and the other sent, on what those notes write — and for each of those: the property's name, a type for it, up to 12 of the distinct values in it, each cut to the first 80 characters with a mark showing it was cut, how many notes hold each of those values, and how many notes the vault has and how many of them hold a value. The vault itself travels as a single opaque marker rather than its name or folder path, and the same workspace identifier and routing details described above accompany this request too. This goes through the same Requesty EU gateway and the same
policy/auroradocspolicy as built-in AuroraAI chat. That type is the one the importer guessed where the notes holding the property agree on it; where they disagree it is a placeholder, sent together with a marker recording that they disagreed, so the fact that a property's values are inconsistent is itself disclosed. The text of your notes is never sent. At most 120 such properties are sent for one import. If you decline the option, if no provider is available, or if the published policy does not yet cover this review, the import keeps the types it guessed on its own and nothing is sent. If you choose your own provider key instead, your device sends that request directly to that provider. - Built-in AuroraAI chat: prompts, retrieved context, workspace routing metadata, server-selected model details, and usage records are sent to Requesty's EU OpenAI-compatible gateway when you use built-in AuroraAI chat. The configured
policy/auroradocspolicy is restricted to approved EU-region providers/models with zero prompt/response retention and no training on requests or responses. Under this configured policy, AuroraAI chat content stays within the EU. Requesty may retain limited operational telemetry (such as token counts, cost, latency, model/provider, status, timestamps, and request IDs), but not prompt/response content. With Aurora's current Requesty configuration, built-in hosted transcription and speech generation are unavailable: those audio routes fail closed instead of sending audio outside the approved chat policy. If you select a BYOK speech provider, your device sends that request directly to your chosen provider. AuroraAI remains unavailable for non-cloud and E2EE-enabled workspaces unless the separate eligibility and consent gates allow it. Aurora does not use your prompts, retrieved context, or outputs to train Aurora-owned models. - Billing and invoices, if paid plans are enabled later: billing contact details, invoice records, tax metadata, payment tokens, and usage counters may be processed by the selected billing provider. Invoice and accounting records follow the billing retention policy and applicable legal obligations once paid plans launch.
3. Purpose and Legal Basis
| Purpose | Legal basis |
|---|---|
| Provide and maintain the service | Contract (GDPR art. 6(1)(b)) |
| Send email notifications (optional) | Consent (GDPR art. 6(1)(a)) |
| Security and abuse prevention | Legitimate interest (GDPR art. 6(1)(f)) |
| Comply with legal obligations | Legal obligation (GDPR art. 6(1)(c)) |
4. Storage and Retention
Cloud-synced data is stored on AuroraCloud infrastructure that we operate ourselves. Our servers are located in Norway (EEA) for the application and database, and Finland (EU) for uploaded files and encrypted offsite backups. The providers that host this data are listed in section 6. Planned hosting changes: none planned. We will notify users before any future move. We retain account data for as long as your account is active. After account deletion, all personal data is removed within 30 days, except anonymised log data which may be retained for up to 90 days for security purposes.
If paid plans are enabled later, billing and invoice records follow the billing retention policy and applicable legal obligations rather than the standard account deletion window.
5. End-to-End Encryption (E2EE)
Aurora supports end-to-end encryption (E2EE) for the main object content body, object titles, property values, comments, workspace chat messages, and uploaded file/image bytes. When E2EE is active, that data is encrypted on your device before it reaches our servers. Your encryption keys are not derived from your login password. Each device you approve holds its own key slot, and your 12-word recovery phrase can restore access if you lose every device. That is why you can sign in with a password, Google, or GitHub without changing how your content is encrypted — and why we cannot unlock your data for you. Save your recovery phrase.
Current scope note: E2EE still does not yet cover all structural metadata. Object type, parent/child relationships, some workspace-level settings, and other server-visible sync metadata may still be readable by the server and are used to power sidebar labels, graph structure, sharing, and similar features.
This means the current implementation should be understood as broad content-and-attachment encryption, not full-workspace encryption.
- The encrypted content is produced client-side before cloud sync.
- Save your recovery phrase — we cannot recover your data without it.
- Workspaces without E2EE enabled store content in plaintext on the server, protected by access controls and server-side isolation rules.
6. Third-Party Processors
We do not share personal data with third parties, except:
- Gigahost AS — server hosting for the application, API and database (Norway, EEA). Gigahost acts as a data processor under a data processing agreement.
- Hetzner Online GmbH — object storage for uploaded files and encrypted offsite backups (EU region, Finland). Hetzner acts as a data processor under a data processing agreement. We back up the database every day. Each night, backups of the database and of uploaded files are encrypted and copied to Hetzner in Finland, where they are kept for 30 days and then deleted, so data you delete can remain in a backup for up to 30 days.
- Resend Inc. — email delivery, only when you have enabled email notifications. Message dispatch runs on Resend’s EU infrastructure in Ireland. Resend Inc. is incorporated in the United States, and account data, delivery metadata, logs, and API records are processed there under a data processing agreement with standard contractual clauses. Resend acts as a processor only when email notifications are enabled.
- Requesty AI — built-in AuroraAI hosted chat inference, the optional hosted review of imported database column types, and the optional hosted review of imported Obsidian vault property types, only when you use those features (EU gateway, approved EU-region providers only). AuroraAI chat runs only on EU-region providers configured for zero prompt/response retention, under the
policy/auroradocspolicy, with no training on requests or responses. Requesty may retain limited operational telemetry such as token counts, cost, latency, model, status, timestamps, and request IDs, but never prompt or response content. Hosted transcription and speech generation are not sent through Requesty. - Staan — EU web search for Research Studio and the assistant's web search tool, only when you use those features (European Union, OVHcloud infrastructure). Staan is Qwant/Ecosia's European Search Perspective product: its own search index, hosted in the EU. Only your search queries and request metadata are sent; page reads keep going through our own self-hosted web fetcher, not Staan. Staan offers an opt-in American backup mechanism that could otherwise use some results; it is disabled for our account. A data processing agreement with Staan is in place. Tavily, a separate United States-based web search provider, stays disabled and unconfigured because it is not EU-only.
- Google LLC — Google sign-in, and the optional Google Calendar connection (requests are relayed by our servers in Norway, EEA). Google is the source of the data you choose to bring in, not a processor that stores your Aurora content. It receives the sign-in and calendar requests you make, and Google handles them under its own terms. This applies only if you sign in with Google or connect Google Calendar; what Aurora reads and where it is kept is described in section 2.
We never sell personal data.
Aurora's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
7. Your Rights
Under GDPR you have the right to:
- Request access to the data we hold about you.
- Request correction of inaccurate data.
- Request deletion ("right to be forgotten") — available directly in Settings → Account.
- Withdraw consent (e.g. email notifications) at any time.
- Lodge a complaint with your local data protection authority.
8. Data Processing Agreement (DPA)
Business customers who process personal data through Aurora may request a data processing agreement by contacting us at contact@aurora.ink.
We keep a current subprocessor register for business customers. The active processors are the ones listed in section 6. Business customers can request the latest subprocessor list and a DPA copy by emailing contact@aurora.ink.
9. Cookies and Local Storage
Aurora uses no third-party tracking cookies. We use the browser's localStorage, sessionStorage, and IndexedDB to store user preferences, offline content cache, folder sync handles, and short-lived handoff state locally on your device. This notice is informational only. This data never leaves your device unless you explicitly enable cloud sync.
Public website statistics
On our public marketing website, Ogard Labs uses self-hosted Plausible at analytics.ogard.cloud to understand visits and improve these pages. This is separate from the signed-in Aurora application and its optional diagnostics.
We count visits only to the home, features, download, privacy and terms pages. We send the public page path without query strings or fragments. We do not send document content, account identifiers, form values or referring URLs. The receiving server processes your IP address and browser information to derive visitor statistics, such as browser and country. This measurement uses no cookies or local storage and respects Do Not Track and Global Privacy Control.
Statistics are hosted on our infrastructure in Norway. Historical statistics are retained until manually removed; no automatic expiry period is currently configured. Contact contact@aurora.ink with questions or objections. This website notice does not enable tracking inside the application.
10. Changes to This Policy
We will notify you by email of material changes. Minor updates will be published on this page with an updated date.