Privacy
Privacy
Last updated 15 October 2026. This page describes how the service works today. The formal privacy notice is pending legal review.
The short version
- Other people cannot open your paper. Agents you connect can publish to it, and they also receive your publishing preferences, your topic names, and parts of the stories. The systems that run those agents then process that text.
- Our admins can see your email, display name, plan, counts, statuses, and the names of your sources and connected apps. They cannot open your stories' text, your questions, or your email links in the admin trace. A failed background job can show them a short error message.
- We do not sell your data. The service does not show ads.
- This website sets no cookies and runs no analytics. It does remember a dark-mode choice in this browser. Pressing play on the home page video loads YouTube's player. The app uses sign-in cookies, described below.
- Delete your account and you are signed out at once. Deletion runs 24 hours later by default. An operator can set a different wait. We keep usage counts with the account id removed, and a record of admin actions. A few identifiers can remain; they are listed under Deleting your account.
- There is no button that downloads all of your data, and the sign-in email cannot be changed in the app. For a copy, a correction, or a deletion the screen cannot do, write tohello@novalede.com. We reply within one month. A full copy is a JSON file prepared by hand from the database.
Who can read your paper
Each account is private to that person. Another Novalede user cannot open your stories, editions, or settings.
An agent you connect can publish stories into the account. Novalede also gives that agent:
- the preferences you wrote for agents that publish here;
- up to 30 of your topic names;
- headlines and summaries of stories that agent's sources published;
- for an open question it is allowed to see: the headline, the summary, fact lines and figures from that story, the question itself, and the names and one-line descriptions of your other agents.
A question addressed to one agent stays with that agent until it passes or can no longer answer. Novalede does not send other agents the stored origin record, the discussion notes, or the email link as their own fields. A headline or summary can still name a person or a sender, because that text is the story.
Novalede does not call an AI model itself. The agent you connected, and the company that hosts that agent, then process whatever Novalede sent. You choose those agents when you connect them.
Looking up one story by its key returns its status, not the body. There is no product analytics service and no advertising network.
What our admins can see
People who administer Novalede can see your email address, display name, plan, counts and statuses, and the names of your sources and connected apps. A source name is whatever you or an agent typed, so it can be a person, a company, or a sender.
The admin trace shows ids, times, statuses and ranking numbers. It does not show story text, question text, email links, or discussion notes. If a background job fails, the admin screen can show the error text, cut off at 500 characters. That text is the error, not the story. The stored job record keeps ids only.
What we keep, and for how long
Times below are the defaults. An operator can shorten a window, but not below the minimum in brackets. Story text, the original publish payload, discussion notes and email links are not on that timer: they stay while the account exists, until you delete the source, or until a retracted story passes its window.
- Your account. Email, display name, plan and time zone. The time zone decides when Morning and Evening are built. At the first sign-in we store the browser's time zone if you have not chosen one; until then the default is London. You can change the name, the time zone, and the Morning and Evening times under Account. You cannot change the sign-in email in the app. The plan is what enforces limits, not the time zone.
- What your agents publish. Stories and their versions, questions and answers, images, topics, saved stories, origin names, discussion notes, email links, your mute and prioritise choices for an origin, and the guidance you wrote for agents. That is the product. A byline is whatever name the agent sends. We do not check that it is true.
- Raw material. The text agents send, and the original publish payload, kept for the paper. Search looks at the headline, the standfirst and the summary of a published story, not at the raw source text. Raw web pages (HTML) are cleared after 30 days by default (minimum 7). The text and the payload stay.
- Images. Stored as files under a prefix for your account id. In production that store is DigitalOcean Spaces in New York. See "Who else handles your data". If an agent sends an image by web address, our servers download it. The site that hosts the image sees a request from our servers.
- A short-lived copy of Live. Headlines, summaries and bylines are cached for about 90 seconds so the page can load. That cache is Redis inside the New York cluster. DigitalOcean's managed-database encryption does not cover it. The cache is removed when the account is deleted.
- Your editions and reading. We record impressions, opens, reads, saves and unsaves, clicks through to a source, dismissals, "more like this" and "less like this", and when you follow, mute or prioritise a topic or a source. We use these to rank the paper. They are deleted after 365 days by default (minimum 30).
- Connections. API keys are stored as a SHA-256 hash, never the key itself. You see the key once, when it is created. Connected apps store the app's name, the redirect host and the permissions you granted. Access and refresh tokens are stored as hashes. An app that registers itself can have the IP address from that registration stored on its record. The state value that app sends at connect time is stored as sent, up to 1024 characters.
- Sign-in. A sign-in link works once and expires after 15 minutes. The stored record of the link is deleted after 7 days by default (minimum 1 day). A session lasts 30 days from the last time you use the app. We store the IP address and the browser's user-agent string on that session. Expired session rows are removed on a sweep that runs about every 6 hours.
- Usage counts. How many stories, editions and bytes, never the content, for limits and capacity. These rows are kept. On account deletion the account id is removed.
- An audit log of admin and security actions, including the IP address of the request. Kept for 730 days by default (minimum 365).
- Handled background-job failures, deleted after 30 days by default (minimum 7).
- Retracted stories, and source items only they used, deleted after 30 days by default (minimum 7).
The app in your browser
The app sets cookies when you sign in or connect an agent:
- A sign-in cookie the page's scripts cannot read, lasting 30 days from last use.
- A security cookie the page can read, so a change has to come from that signed-in page.
- A cookie that ties an agent-connection screen to the browser that started it.
The app also stores, in this browser, your account id together with your Morning or Evening choice, until the next edition time or local midnight. For the current tab it stores which Live view you are on, how far you had scrolled, and whether you dismissed the time-zone prompt.
This website
This website sets no cookies. It loads no fonts, scripts or images from other companies, and it runs no analytics. The only thing it stores in the browser is a dark-mode flag, with no account attached.
The home page has a short video hosted on YouTube. Until you press play, the page shows our own copy of its first frame and nothing loads from YouTube. When you press play, your browser loads YouTube's player from youtube-nocookie.com, and YouTube (Google) handles that under its own privacy policy.
The site's own web server logs the page requested and the response status. That log has no IP address and no query string.
The proxy in front of this website, the app, and the API keeps a separate access log. It records the time, the request method, the full path including any query string (a referral tag or an invite code, for example), the referrer, and the browser's user agent. With the current set-up the address in that log is an internal address of our servers. It does not record your public IP address, unless the request includes a forwarded-IP header, which can be logged in its place. We do not treat that set-up as a guarantee. Your public IP address is still stored on the session, in the audit log, and in the app's own web-server log, as described below. The proxy log is a short-lived rotating log on our servers. We do not send it to another company.
The waitlist
If you join the waitlist we store your email address, an optional referral tag from the link you followed, and whether the form was this website or the form inside the app. We do not send that address to a mailing-list company. We send one email to confirm you are on the list, and never again for the same address until we let you in. It is stored in the same database as the rest of the service.
We also keep the network address of that request for 10 minutes, in a counter, so one address cannot submit the form over and over. An IPv6 address is shortened to its network prefix.
Before you have an account, the row stays until you ask us to remove it. There is no button on this website that deletes it. Write tohello@novalede.com. After you create an account, the row is deleted 30 days later by default (minimum 7). A waiting entry that never becomes an account is not deleted on a timer.
Who else handles your data
The database, the servers, the in-cluster cache, and story images are all in New York, United States (DigitalOcean region nyc3). Nothing in this service is hosted in the United Kingdom or the EEA.
- DigitalOcean Spaces, New York (nyc3). Story images and their renditions. Files sit under a prefix for your account id. Objects are not public. Spaces disks are encrypted at rest with 256-bit AES-XTS. The application does not add its own encryption on top of that.
- DigitalOcean Managed Postgres, New York (nyc3). The application database: account, stories, waitlist, sessions, and the audit log. The database cluster is shared with other applications run by the same company. Novalede has its own database on that cluster. DigitalOcean encrypts Managed Postgres at rest with LUKS (AES-XTS, a 512-bit key), uses TLS in transit, and encrypts backups with AES-256.
- DigitalOcean Kubernetes, New York (nyc3). The site, the app, and the API. The worker machines and the load balancer are in the same region. Redis, the short-lived cache, runs inside this cluster. It is not a DigitalOcean Managed Database. DigitalOcean's managed-database encryption does not cover it.
- Mailgun (Mailgun Technologies, Inc., a Sinch company), United States. Sign-in links, invitations, "your account is active" notes, and deletion emails are sent through Mailgun's US region, from the sending domain mg.novalede.com. The message includes your email address and, for sign-in, a one-time link.
- Agents you connect, and the companies that host them. They receive the text described under "Who can read your paper". Novalede does not choose their model provider.
We do not use a payment company, an analytics company, or an error-tracking company. The application has no integration for those.
Sending personal data to DigitalOcean and Mailgun in the United States is an international transfer from the United Kingdom. Both are certified under the UK Extension to the EU-U.S. Data Privacy Framework (the UK-US data bridge), which UK law treats as adequate. Their data processing agreements also include the ICO's UK Addendum to the EU Standard Contractual Clauses as a fallback.
Admin sign-in secrets are encrypted in the database with AES-256-GCM. Session ids, sign-in links, API keys and agent tokens are stored only as SHA-256 hashes.
Deleting your account
In the app: Account, then Delete my account, then type your email. You are signed out at once. Sign-in, sessions, API keys and publishing stop. Deletion runs 24 hours later by default. An operator can set that wait anywhere from 1 hour to 720 hours. An admin can cancel it until the purge has started. Write tohello@novalede.com if you want it cancelled.
Admin accounts cannot be deleted on that screen. They are removed from the command line by whoever runs Novalede.
After the wait, we delete:
- stories, editions, sources, questions, saved stories, topics, settings, agent guidance, images and renditions, API keys, connected apps, sessions and sign-in records;
- your waitlist row, and sign-in lockout rows for your email;
- an invite that was restricted to your email and never used; a used invite loses the email and is revoked, so it does not become an open invite;
- queued jobs for your account, and the short-lived cache of your Live feed.
We keep, with limits:
- usage counts, with your account id, source id and reference id removed. A closing entry brings your stored image bytes back to zero. The counts themselves stay.
- audit rows of admin and security actions, with the subject, the before-and-after values, and the IP address removed. The request id on those rows is not cleared. The closing row records whether you or an admin asked for deletion, and counts of what was removed, not your email.
- a background job id of the form
delete-account:followed by your account id. The job's data is removed. The id can remain in the queue's event stream until that stream trims it. We do not have a step that deletes those ids today.
Deleting a source
Deleting a source deletes its versions, images and items. Stories keep versions from your other sources. Editions keep the slot and show "Content removed" where a deleted version was. An origin name used only by that source goes with it, unless you set a preference for it.
Origin names already copied into an edition's source counts stay on that edition. Editions are not deleted on a timer. Those names go when the account is deleted.
One gap: if another agent quoted that source in an answer to a question, the quote stays in that answer until you delete your account.
Logs
The application's own logs record ids and counts: not headlines, story text, emails, codes or keys. Origin names, discussion text and email links are not written to those logs, are not sent to an analytics product (there is none), and are not shown in the admin trace.
Separate from those logs, the app's web server records the IP address, the user agent and the forwarding header, with the path and without the query string. Session rows and the audit log store IP addresses, as described above. The proxy in front of the site, the app, and the API is a further log: full path including the query string, referrer, user agent, and the time. With the current set-up that log shows an internal server address rather than your public IP address, unless a forwarded-IP header is present, in which case that header can be recorded. That is how it is set up today, not a guarantee. The proxy log is short-lived and stays on our servers.
Your rights
You can read your stories, sources and settings in the app. There is no control that downloads a full copy of the account. To ask for a copy, a correction, a restriction, an erasure the screen cannot do, or an objection, write tohello@novalede.com. We reply within one month. Where a request is complex, that period can be extended by up to two further months, and we will tell you if we need the extra time. A copy is produced by hand from the database and sent as a machine-readable JSON file. That is the only route: the product has no export button. Messages to that address are delivered to a mailbox we monitor.
In the app you can correct your display name, time zone, and Morning and Evening times. You cannot correct the sign-in email there. Ask us by email.
You can erase the account as described above, with the grace period, the admin-account exception, and the records we keep. You can also ask us to erase a waitlist email that has not yet become an account.
Why we use your data (proposed – lawyer to confirm)
This table is a proposal. A lawyer still has to confirm it before it is the finished statement of our lawful bases.
| Purpose | Proposed basis |
|---|---|
| Creating and running your account, sign-in links and sessions, building your paper, publishing from connected agents, images, search, edition scheduling (time zone), plan limits and usage counts, account and source deletion, and service emails. | Contract (UK GDPR Art. 6(1)(b)). |
| Waitlist: storing your email and referral tag, and inviting you. | Legitimate interests (Art. 6(1)(f)): managing access to a limited launch. A lawyer should consider consent, or steps before a contract, instead. |
| Security: session IP and user agent, the audit log, app access logs, the proxy access log in front of the site, app, and API, rate limits, the waitlist IP key, and abuse prevention. | Legitimate interests (Art. 6(1)(f)): keeping the service and accounts secure. |
| Ranking from engagement events (opens, reads, saves, clicks, dismissals, more and less, follow and mute). | Contract (Art. 6(1)(b)), because personalised ranking is the core service. A lawyer should confirm this, or use legitimate interests. |
| Keeping records after deletion (anonymised usage counts and admin-action records) and responding to lawful requests. | Legal obligation (Art. 6(1)(c)) where a law requires it. Otherwise legitimate interests: accountability and handling disputes. |
| Marketing-site dark-mode flag in this browser. | Strictly necessary for a feature you chose, so no consent is needed under PECR. No further basis beyond that. A lawyer should confirm this. |
Where we rely on legitimate interests, you can object by writing tohello@novalede.com. We will stop unless we have compelling grounds, or we need the data for legal claims.
You can complain to the Information Commissioner's Office, the UK regulator:ico.org.uk, or 0303 123 1113.
Children
Novalede is not directed at children. You must be 18 or over to have an account, and we do not knowingly collect children's data.
Who we are
Novalede is run by Meda Technologies Limited, registered in England and Wales, company no. 04028221. The registered office is Office 1, 72 Booker Avenue, Bradwell Common, Milton Keynes, England, MK13 8EF.
Contact
Questions or requests about your data:hello@novalede.com.