Last updated 29 July 2026
This document covers SOYL Labs’ hotel products. Policies for this website, Quill and checkout are being finalised with our lawyers; until then, write to outreach@soyl.cloud with any question.
1. Who we are
SOYL AI is operated from Bengaluru Karnataka 560043, India. For anything in this policy, write to ryan.gomez@soyl.cloud.
This policy is written under the Digital Personal Data Protection Act, 2023 (“DPDP”). Where you are in a jurisdiction with stronger rights, we apply those.
2. What we collect
Account data. Your email address, a display name if you give one, and a hash of your password. We never store the password itself.
Workspace data. The name of your property or group, the properties you add, and who else you invite.
Documents you upload. The file, the text extracted from it, and numerical representations of that text used for search. See section 4.
Questions and answers. Every question asked in your workspace, the passages retrieved to answer it, and the answer given. This is kept permanently while your account is open — it is how the product is improved, and how we can explain why a particular answer was given.
Technical data. IP address, browser and device type, and timestamps, recorded when you use the service.
We do not use cookies for advertising or tracking. The only cookie we set is the one that keeps you signed in.
3. What we do with it
Account and workspace data operates the service: signing you in, showing you your own data and nobody else’s, and sending the two emails we send — address verification and password reset.
Documents are used for one purpose: answering questions asked inside your own workspace. They are not used to train any model, ours or anyone else’s. They are not shared with other customers. They are not read by us except where you ask us to look at something specific, or where we are compelled by law.
Questions and answers are used to measure and improve retrieval quality, and to decide what to build. Where we use this to improve the product, we work with the question text and our own performance data — not with the contents of your documents.
4. Your documents, specifically
This is the part most people want to understand, so it is stated plainly.
When you upload a document, the file is stored encrypted. We extract its text, split it into passages, and compute an embedding for each — a list of numbers representing meaning, which is what makes search work. The file, the passages and the embeddings all live in your workspace and are readable only within it. This is enforced in the database itself rather than in application code, so a mistake in our software cannot expose one customer’s documents to another.
When you ask a question, the passages most relevant to it — and only those — are sent to our model provider along with your question. Whole documents are never sent.
Before that happens, we strip out patterns that look like guest personal data (names, email addresses, phone numbers) where we can identify them, so that the minimum necessary content reaches the model.
5. Who else processes it
We use the following sub-processors. This list is current as of the date at the top of this page, and we will update it here before adding a new one.
| Processor | Purpose | What reaches them |
|---|---|---|
| Microsoft Azure (Azure OpenAI Service) | Generating answers, embeddings and search suggestions | Passages from your uploaded documents that are relevant to a question, and the question itself |
| Railway | Application hosting, database and job queue | All account and document metadata, document text, and the answer history |
| Cloudflare R2 | Storage of the document files you upload | The original files, exactly as uploaded |
| Vercel | Hosting and delivery of the website and application interface | Request metadata such as IP address and user agent |
| Resend | Transactional email — verification and password reset only | Your email address and the contents of those messages |
| Plausible Analytics | Website analytics | Page views and funnel events. No cookies, no cross-site identifiers, no personal data |
One disclosure worth reading carefully. Microsoft retains prompts and the model's responses for up to thirty days for abuse monitoring, and may review them where automated systems flag a potential policy violation. Microsoft does not use this content to train its models. We have chosen to leave abuse monitoring enabled because it is currently the only control watching for misuse of the model, and we will disclose here if that changes.
6. How long we keep things
Documents are kept until you delete them or close your account.
Questions and answers are kept while your account is open, because the history is part of the product.
Access and audit records are kept for twelve months. These record who did what, and shortening them would remove our ability to investigate a security incident.
On account closure, everything above is deleted within thirty days, other than records we are required by law to retain.
7. Deletion means deletion
When you delete a document, we remove the file, the extracted text, every passage derived from it and every embedding. It stops being findable immediately and is gone rather than flagged as hidden.
Encrypted backups may retain a copy for up to thirty days, after which they expire. We do not restore deleted content from backup.
8. Your rights
Under the DPDP Act you may:
- ask what personal data we hold about you and get a copy;
- have inaccurate data corrected;
- have your data erased;
- nominate someone to exercise these rights if you are unable to;
- complain to the Data Protection Board of India.
Write to ryan.gomez@soyl.cloud and we will respond within thirty days. We will not charge you for a request, and we will not ask why you are making one.
If you are an employee of a hotel using SOYL, your employer is the data fiduciary for the documents in their workspace and we act on their instructions — address requests about that content to them, and we will help them respond.
9. Security
Data is encrypted in transit and at rest. Access to production systems is limited to the people who need it and is logged. Passwords are hashed with Argon2id. Separation between customers is enforced by the database, and we run an automated test suite against that separation which cannot be skipped before a release.
If we discover a breach affecting your data, we will tell you and the Data Protection Board without undue delay.
10. Children
SOYL is a tool for hotel staff and is not directed at anyone under 18. We do not knowingly collect data about children.
11. Changes
If we change this policy in a way that materially affects you, we will email you before it takes effect. Minor corrections are made here with the date at the top updated.
See also our Terms of Service and, if you are a customer, Data Processing Addendum.
Also see terms of service and data processing addendum.