What this service collects, and what it does not
Written to describe what the software actually does. If the two ever disagree, the software is the bug.
The short version
You can search and read the entire statute book without an account and without giving anything. Saving a provision, revising, practising or holding a conversation stores that work against an identifier so it can be shown back to you. Uploaded files are never stored; what Saul writes from them is saved in your conversation, like every answer. You can export everything held about you, or delete all of it, at any time.
Who is responsible
This service is operated by an individual, not a company, and is currently self-hosted. For any question about your data, or to exercise the rights below, use the details on the contact page.
What is held
| Data | Why | When |
|---|---|---|
| A random identifier stored in your browser | Keeps your saved work separate from everyone else's | The moment you first save, revise or practise |
| Email address, name and profile picture | So an account survives clearing your browser, and works on a second device | Only if you choose to sign in with Google, which supplies them |
| Email address and a one-way hash of your password (scrypt, salted per account) | So you can sign in without a Google account | Only if you create an account with an email and password |
| Saved provisions, notes, flashcard review history, practice scores | They are the feature | When you create them |
| Conversation threads, including your questions and the answers | So a follow-up question can refer to what came before | When you use a conversation rather than one-off questions |
| Messages you send from Help & support, the email you give for a reply, and your browser type; and, only if you tick the box, your last question and the provisions it showed | So the problem can be understood and answered | Only when you send one |
| Voice input. If you use the microphone button, your browser's own speech service turns what you say into text (in Chrome, that is Google's). This service receives only the text, as if you had typed it | So you can ask by speaking | Only when you press the microphone button |
| Your rating of an answer or case note. "Helpful" keeps the rating alone; "Not helpful" also keeps your question (for a case note, your file's name, never the file), the answer or note, the provisions it cited, and anything you write | So a wrong answer can be found, checked and fixed | Only when you rate one |
| Payment records: order reference, amount, date, status | Required to know what was paid, and for accounting | Only if you pay |
| Ordinary web server logs, which include IP addresses | Keeping the service running and diagnosing faults | Every request, as with any website |
What is never held
- Uploaded files. A file you upload for a case note is read in memory and discarded: neither the file nor its text is written to disk. The note or answer written from it is saved in your conversation like any other, and may quote it.
- Card, UPI or bank details. Payment is handled entirely by Razorpay. Those details go from your browser to them and never reach this server.
- Your password itself. If you sign in with an email and password, only a salted scrypt hash is stored, which cannot be turned back into the password. If you sign in with Google, no password reaches this server at all, and nothing from your Google account is kept beyond the email, name and picture listed above.
Who else sees your questions
This is the part most services are vague about, so plainly: generating an explanation means sending your question, and the statutory text retrieved for it, to a third-party language model provider. Analysing a case also sends the text of the document you upload — up to about 40,000 characters of it, the beginning and end of a longer one — so the model can identify the issues and write the note. The provider processes it under its own terms. Searching, browsing and reading statutory text involve no such call and reach no third party.
- OpenAI — runs the language model. Under OpenAI's published API policy, it does not use what it is sent to train its models, and keeps it for up to 30 days to monitor for abuse before deleting it.
- Groq — runs the language model. It processes requests under Groq's own published terms. Used only when OpenAI cannot be reached.
- Google — only if you choose to sign in with Google, and only to confirm who you are.
- Razorpay — only if you pay, and receives the payment details directly from you.
- Cloudflare — carries traffic to this service and therefore sees requests in transit.
Your data is not sold, and is not shared with anyone else for advertising or any other purpose.
Your rights under the DPDP Act 2023
The Digital Personal Data Protection Act 2023 gives you a right to access what is held about you and to have it erased. Both are built in rather than being a request you have to make and wait on:
- See what is held — Settings, under Your data, shows a summary.
- Download all of it — as a JSON file, from the same place.
- Delete all of it — one action, immediate, irreversible.
Deletion removes your saved provisions, notes, review history, practice scores, conversations, documents and account. Backups made before you delete keep a copy until they expire, within 120 days; they are not part of the live service. If one of them is ever restored, your deletion is applied to it again before the service is used. A short record of the deletion is kept for that purpose, on the server and in storage separate from it: it holds no name, email or content. Nothing is deleted until that record is stored safely off the server. If it cannot be, you are told the deletion did not happen, nothing is removed, and you can try again. Payment records are the one exception: they are anonymised rather than deleted, because a completed transaction carries its own retention obligations. The amount and date survive with no link to you. So does what each answer cost to write, in tokens and rupees, with no link to you or to what was asked: it is how the plans' usage limits are checked.
How long things are kept
Your saved work is kept until you delete it or delete your account. There is no automatic expiry, because a revision schedule that quietly erased itself would be worse than useless. Answers are saved in your conversations whether or not the file they were written from was kept. Backups keep a copy of everything for a while after it is deleted, within 120 days. Server logs rotate in the ordinary way, and never carry the text of your files. Anonymised payment records are retained as long as accounting requires.
Children
This service is built for people practising or studying Indian law and is not intended for anyone under 18. The DPDP Act imposes additional obligations for children's data that this service is not currently set up to meet.
Security, stated honestly
Traffic is encrypted in transit. Sign-in sessions are httpOnly cookies stored as hashes, so a copy of the database yields no usable sessions, and passwords are held only as salted scrypt hashes. Against that: this is self-hosted on ordinary hardware by one person, without a security team, formal audit or breach-monitoring. Judge what you upload accordingly — and note that uploaded files are not stored at all, which is the strongest protection available for them; what is written from them is kept in your conversations.
Not legal advice
Everything this service produces is research, not advice. Statutory text reflects what India Code published when it was collected and may not include the latest amendment. Verify against the official source before relying on it.
Changes
If what the software does changes, this page changes with it. There is no version history yet; when there is a userbase to notify, there will be.