BanglaLLMBanglaLLM logo mark
Sign inJoin in

Legal

Privacy Policy

BanglaLLM is a free volunteer project that collects writing and voice recordings in Bangladeshi Bangla and its dialects. This page explains, in plain words, what we collect about you and what we do with it.

Last updated: 18 November 2026

Contents

  1. What we collect
  2. Why we use it
  3. Who can see your information
  4. The dataset and open publication
  5. Withdrawing your consent and deleting your data
  6. Cookies and analytics
  7. How long we keep it, and security
  8. Children and young people
  9. Changes to this policy

What we collect

When you register and take part we collect the following.

  • Account details: your name, a one-way hash of your email address (used to sign you in, not the address itself), and a password kept only in scrambled (hashed) form.
  • About you: your age range, gender, home division and district, and the dialects you speak. We ask for these so the data can be sorted by region. We never ask for your exact age, address or phone number.
  • Your contributions: the answers you write, the voice recordings you make when voice tasks are open, and the corrections you give.
  • Your reviews: your votes and notes when you check other people's work.
  • Your consent: which wording you agreed to and when.
  • Technical details: basic information about your browser and device sent with a submission (such as the type of device and how long a recording is), and security logs such as sign-in attempts.

A recording of your voice can identify you, so we treat it as personal data.

Why we use it

  • To build a dataset for fine-tuning existing open weight models (for example Gemma and Qwen) so they understand, and speak, Bangladeshi Bangla and its regional varieties better. We do not train a new model from scratch.
  • To check quality: other participants from your region, and project staff, review contributions.
  • To run the service: signing you in, sending sign-in codes by email, showing your progress and keeping the platform secure.
  • To understand how the site is used, in aggregate (see Cookies and analytics below).

We do not use your voice to imitate you personally, we do not sell your data, and we do not use it for advertising.

Who can see your information

  • Other participants see only your random ID (like BL-7K3M-9QXA) and the avatar you chose. They never see your name.
  • Your name is visible only to you and the project administrators.
  • The public pages show only totals and counts, and small groups are hidden so nobody can be picked out.

Service providers help us run the platform (for example hosting, email delivery and traffic delivery). They handle data only to provide that service. If AI screening of text is switched on, the text you wrote may be sent to an AI service for a first quality check, and we will ask you to agree again when that happens.

The dataset and open publication

Accepted contributions are collected into a dataset. The dataset may be used for research and may be published openly so others can improve Bangla language technology. Published records carry a random ID and the information needed for research, such as dialect and region, and never your name or email.

Once a dataset has been published and copied by others, we cannot take every copy back. That is why we ask for your clear consent before you contribute.

Withdrawing your consent and deleting your data

You can withdraw your consent at any time from your Account page. When you do:

  • your voice recordings are deleted;
  • your answers are removed from all later datasets;
  • your name is deleted.

Datasets that were already published before you withdrew cannot be recalled. You can also ask us, using the contact details below, to see, correct or delete the information we hold about you.

Cookies and analytics

  • A sign-in cookie keeps you signed in. It is needed for the service to work.
  • A language cookie remembers whether you chose English or Bangla.
  • We count visits to the public pages (the home, sign-in, register, privacy and terms pages) with a small counter of our own. It sets no cookie and does not store your IP address or any account information. It keeps only the page, the site you came from, your device type and language, and a code that changes every day and cannot be traced back to you. It is off when your browser sends Do Not Track.
  • Messages and feedback you send us from the portal are stored with your participant ID so we can answer them. They are deleted when you withdraw your consent.
  • On the live site we use Google Tag Manager to measure, in aggregate, how the public pages are used. We do not use it for advertising.

You can block or delete cookies in your browser settings, but signing in will not work without the sign-in cookie.

How long we keep it, and security

We keep your account and contributions while the project runs, or until you withdraw your consent or ask us to delete them. Security logs are kept for a limited time.

We protect data with access limits, hashed passwords and email addresses, and encrypted connections. No system is perfectly secure, so we cannot promise that nothing will ever go wrong, but we work to keep your information safe.

Children and young people

Taking part is open to people of all ages, but if you are under 18 please take part only with the agreement of a parent or guardian. If you believe a child has joined without that agreement, contact us and we will remove their information.

Changes to this policy

If we change how we use your information in a way that matters, we will update this page and the date at the top, and where the law or fairness requires it we will ask for your consent again.

Contact

Questions, requests or concerns about this page? Contact the BanglaLLM project team through the contact details given when you were invited to take part.

Terms and Conditions · Back to the home page

BanglaLLMBanglaLLM logo mark

My language, my AI

A data-collection platform. Your writing and voice recordings become a dataset that helps existing open weight models (for example Gemma and Qwen) understand and speak Bangladeshi Bangla. We are not building a new model from scratch.

The project

AboutHow it worksDialectsFAQ

Take part

Join inSign inSee an example task

Legal

Privacy PolicyTerms and Conditions
© 2026 BanglaLLM. A free, volunteer project.