Privacy Policy

AkiraYukti Compute (compute.akirayukti.com) · Effective 13 November 2026

Experimental research service. Data you submit is processed on devices that other people own, and the service may lose data. Do not submit anything confidential, sensitive or personal, and use the pool at your own risk (see the Terms of Use).

This policy explains what AkiraYukti Compute (the "pool") stores and why. The pool is run by AkiraYukti, the research division of ChaiSaab; in this policy "we" means AkiraYukti. Our research home page, akirayukti.com, stores no personal data. ChaiSaab's own website and services have their own policies.

Short version: we store as little as the pool needs, we do not sell your data, and there are no cookies, advertising trackers or analytics.

1. What the pool stores

Data	Why
Email address	To sign you in with a one-time code and to identify your account. Sign-in codes are stored only as a hash and expire after 10 minutes.
IP address	To limit sign-in attempts and requests, to block abuse (abuse bans cover the email and the IP addresses used), and in our security log.
Account and usage records	Your quota, jobs, results, models and datasets you create, credits earned and gifted, and the times of these actions. Needed to run the service and keep credits correct.
Content you submit	Datasets, files, links to datasets, text, prompts, chat messages, agent flows and other job inputs, and their results. Needed to run your jobs.
Device information (if you contribute a device)	Processor cores, memory, graphics capability and name, measured speed, memory budget, current load and the tasks it runs. Needed to schedule work on it. Other users see your device only as "Device N" with its capabilities; never your name, email or IP address.
Security log	An append-only record of sign-ins, failed attempts, account changes, bans, dataset fetches and similar events, with account ids and sometimes IP addresses. Used to investigate abuse and errors.

We do not use cookies, advertising, analytics or third-party tracking scripts. The app stores your settings and sign-in token in your browser's local storage, separately for each account on the same browser, and you can clear it at any time.

2. Where your data is processed

Our server (a virtual private server rented from Hostinger) stores accounts, jobs, results and logs. Stored files are encrypted at rest (AES-256-GCM), and the keys are readable only by the server's administrator account.
Contributed devices. This is how the service works: parts of your jobs run on devices that other people contribute. A device receives the inputs of the tasks it runs, for example a chunk of your data, rows of your training set, or the internal values of a language model processing your prompt. The device that runs the first layers of a language model can effectively reconstruct your prompt. Contributors are not our employees and we cannot fully control what they do. Do not submit personal data about others, passwords, financial or health information, or anything confidential.
Direct device links. For language-model chats, browser devices on the same network may pass data directly to each other. They then see each other's local network address.
Email delivery. Sign-in codes are sent through our email provider.
Dataset links. If you give a link to a dataset, our server downloads it; the site you link to sees our server's address, not yours.
Certificates. We use Let's Encrypt for HTTPS certificates; no personal data is shared with it.

We do not sell or rent personal data and do not share it with anyone except as described here, or when required by law.

3. Public information

The credit ledger is public so that anyone can verify it. Its entries show random device identifiers, hashed account identifiers, internal account numbers for gifts, amounts and times. They never show emails, names or IP addresses.

4. How long we keep data

Results, models and datasets: deleted automatically 7 days after they are created, unless you choose to keep them longer (up to 30 days).
Your account and email: until you ask us to delete the account.
Abuse bans: expire after 7 days (strikes) or 90 days (serious abuse), after which the ban record is removed.
Server logs rotate automatically and are overwritten as they fill up. The security log and the credit ledger are kept as long as needed for security and verification; ledger entries cannot be removed without breaking the ledger, so on deletion we remove the link between those entries and you.

5. Your rights

You can download everything stored for your account from the app (My account → Download everything). You can ask us to correct your data, to delete your account and data, or to explain how your data is used, by writing to contact@chaisaab.com from the email address of your account. We aim to reply within 30 days. If you are not satisfied, you may complain to the data protection authority where you live (in India, the Data Protection Board of India).

6. Age

The service is for people aged 18 or over. Do not use it if you are younger.

7. Security

We use HTTPS, encryption at rest, one-time sign-in codes, rate limits, a network-free sandbox for tasks on devices and automatic abuse bans. No system is perfectly secure; if we learn of a breach that affects you, we will tell you and the authorities as the law requires.

8. Changes

We may update this policy. We will change the date above, and for significant changes we will show a notice in the app.

9. Contact and grievances

AkiraYukti, the research division of ChaiSaab. Email contact@chaisaab.com for any question, request or complaint about your data.