Privacy Policy
Effective · last updated
This Policy explains what Holograph PressWorks collects when you use HoloRemedi, how your manuscripts are processed, and who else can see them. It forms part of our Terms of Service.
The short version: your manuscripts are processed on infrastructure we operate, with one exception — images that need a description are sent to Azure OpenAI, a Microsoft service. Those requests carry the image plus up to 800 characters of the text around it on the page — never the rest of the book. Section 3 is the detail, and it is the section we expect you to check.
1. What we collect
Account information. Your email address and a user identifier from Microsoft Entra External ID, which handles sign-in. You can sign in with any email address and a password you choose, or with a Google account — either way we never see your password. If you buy credit, Stripe collects your billing address and card details and gives us the address, the last four digits and the card brand.
Content you upload. The EPUB files you submit, the workspace we expand them into, the audit findings, the alt text generated or written, and the remediated files and reports we produce.
Usage data. Page and image counts per book (these determine what you are charged), job timings and outcomes, error traces, and request logs including IP address. We do not use advertising or cross-site tracking.
2. How we use it
- To run the remediation you asked for, and to show you the result.
- To meter and bill usage.
- To diagnose failures — a job that crashes is investigated using its logs, and sometimes its files.
- To send transactional email: receipts, job outcomes, account notices.
- To answer support messages and feedback you send us through the in-app Help button.
We do not use your content to train, fine-tune or evaluate any model, ours or anyone else’s.
3. How your content is processed
Audit, repair, metadata editing, EPUBCheck and export all run on infrastructure operated by Holograph PressWorks in our own Azure subscription. Your files are transferred over TLS, stored encrypted at rest, and reachable only by your authenticated account and, for support, by our staff. Preview links to your assets are signed and expire after one hour.
There is one exception, and it is alt text. Images that need a description are sent to Azure OpenAI, a service operated by Microsoft, together with a short instruction and any caption we found near the image. What is sent and what is not:
- Sent: individual images needing a description, scaled down; a short instruction written by us; and up to 800 characters of the text immediately around that image on its page. Where a figure has a caption that is what travels; most books we see have none, in which case it is the surrounding sentences.
- Not sent: the rest of your manuscript, its other files, its metadata, your account details, or anything identifying you. No customer identifier accompanies the request.
We would rather be exact than flattering here. An earlier version of this page said your text was never sent at all, and a later one implied it was usually a caption. Both were wrong. Across the books we have measured, none carried a single caption element, so in practice this is prose from the page — a description written with no idea what the image is doing in the argument is usually a poor one. The 800-character limit is enforced in code and asserted by a test, not merely intended.
Our Azure OpenAI deployment is a regional one in the United States (East US 2), so requests are processed in that region rather than routed to wherever capacity happens to exist. Microsoft states that prompts and completions sent to Azure OpenAI are not available to OpenAI and are not used to train Microsoft or third-party foundation models.
One thing we want to be straight about: Azure OpenAI applies abuse monitoring, under which a request flagged by its automated classifiers may be retained by Microsoft for up to 30 days and reviewed by an authorised Microsoft employee. We have not obtained the exemption from this, which Microsoft calls modified abuse monitoring, so it applies to our requests today.
We previously offered a per-book mode that kept image processing on our own GPU hardware. That hardware is not running at present and the option is not available, so this page describes what actually happens rather than what we would prefer to offer. If on-premises processing is a requirement for you, tell us before uploading anything — it is on our roadmap and we would rather know who needs it.
4. Who else sees your data
These are all of our processors. There are no others.
| Provider | What reaches them | Why |
|---|---|---|
| Vercel Inc. | Request metadata for the pages you visit: IP address, approximate location derived from it, browser and system information, and the URLs requested, which include project identifiers. Not your manuscripts — uploads go directly from your browser to our own infrastructure. Separately, an anonymous count of page views (see the note below), from which project identifiers are removed before they are sent. | Serving the web application, and counting page views |
| Microsoft Azure | All stored content and compute — this is where the Service runs | Hosting, database, file storage |
| Microsoft Entra External ID | Email address, authentication events | Sign-in |
| Azure OpenAI (Microsoft) | Images needing a description, and captions near them. Never your manuscript text. | Alt-text generation |
| Stripe, Inc. | Email, billing address, card details (card numbers go to Stripe directly and never reach us) | Payments |
| Resend | Email address and the content of transactional messages | Sending receipts and notices |
Counting page views. We use Vercel Web Analytics to see how many people visit, which pages they read and which link or search sent them — we are a new product, and that is how we learn which parts of it are worth building out. It is the analytics of the company that already serves every page, so it adds no one to the table above. It sets no cookies: Vercel identifies a visitor by a hash of the incoming request which is discarded after 24 hours, and states that visitors therefore cannot be tracked between different days or different websites. What is recorded is the page, the referring link, the country, and the browser and device type — never your email address, never an account identifier, and never anything that links a visit to you. Before an event is sent we strip the identifiers out of the URL ourselves, so a visit to one of your books is recorded as `/project/[id]/tags` and not as that book. Pages under /admin are not counted at all.
Signing in with Google, if you choose it. The sign-in page offers a Google button alongside email and password. Use it and Google confirms your identity to us and tells us your email address — and Google learns that you signed in to HoloRemedi. Google is not in the table above because it is not acting for us there: it handles your Google credentials on its own behalf, under its own privacy policy. Sign in with an email address and a password instead and Google is never contacted at all. Either way, your manuscripts are never involved in signing in.
We do not share your data with advertising networks, data brokers, cross-site analytics platforms, or any AI provider other than the one named above. We disclose data to law enforcement only where a valid legal process requires it, and will tell you unless we are forbidden from doing so.
5. Where it is processed
The Service runs in Microsoft Azure in the United States: the application and your files in West US 2, and the Azure OpenAI deployment in East US 2. The web front end is served by Vercel Inc. (440 N Barranca Avenue #4133, Covina, CA 91723, United States), which processes request metadata in the United States and other jurisdictions in which it operates. Stripe and Resend process data in the United States and the European Union.
If you are in the EU or UK, transfers out of your region rely on Standard Contractual Clauses. We do not currently offer EU-only processing; if that is a requirement for you, tell us before you upload anything.
6. Your rights
- Access — see what we hold. Your files and reports are downloadable from your workspace at any time.
- Correction — fix inaccurate account data from account settings.
- Deletion — delete an individual book from its project page, or your whole account from account settings. Both are immediate and irreversible.
- Portability — export any remediated book as a standard EPUB plus a report. No proprietary format, no lock-in.
- Objection and restriction — email us and we will stop the processing in question or explain why we cannot.
Email connect@holographpressworks.com for anything the app does not let you do directly. We aim to respond within 30 days. If you are in the EU or UK you may also complain to your national data protection authority.
7. How long we keep things
We would rather describe this accurately than favourably. Your uploaded books and the files we produce are kept until you delete them — there is no automatic expiry on those. The working copy we expand a book into is different: it is swept after 14 idle days to reclaim storage, and rebuilt from your source the next time you open the book.
| Data | Kept for |
|---|---|
| Uploaded books and remediated output | Until you delete the project or your account. No automatic deletion. |
| Expanded working copy of a book | 14 days after you last touch it, then swept. Rebuilt from your source on demand; a copy holding edits newer than its source is never swept. |
| Export bundles in our archive | 90 days, then deleted automatically by a storage lifecycle rule |
| Account record, credit balance, transaction history | Until you delete the account; transaction records are then kept as long as tax and accounting law requires |
| Support messages and feedback you send through the in-app Help button | Until you delete your account. Kept so a bug report is not lost when the person who filed it moves on. |
| Request and error logs | 30 days, then deleted automatically |
Deleting your account removes your account record and the content attached to it. Backups may retain copies for a short period before rotating out.
8. Security
- TLS on every connection; files encrypted at rest.
- Authentication delegated to Microsoft Entra External ID — we hold no passwords.
- Service-to-service access uses managed identities rather than shared keys, so there is no long-lived credential to leak. Our AI provider is authenticated this way and we hold no API key for it.
- Asset preview links are signed and expire within an hour.
- Secrets are held in Azure Key Vault and fetched by the running service, not stored in configuration.
No system is perfect. If you find a vulnerability, email connect@holographpressworks.com and we will respond quickly and credit you if you would like.
9. Children
The Service is for publishing professionals and is not directed at children. We do not knowingly collect data from anyone under 18. If you believe we have, email us and we will delete it.
10. Changes to this Policy
We will update this page when our processing changes, and the date at the top will change with it. For material changes — a new processor, a new category of data, a new purpose — we will email active users at least 30 days in advance.
11. Contact
Holograph PressWorks — connect@holographpressworks.com. Write to us with "privacy" in the subject and it will reach the right person.
Questions about this document: connect@holographpressworks.com