the Bamya Call landing pages at call.bamya.co. Version: 1.0 Effective: 9 August 2026 Last updated: 17 September 2026
1. Who we are
Bamya Call is a live transcription and call-assistance service operated by ATTN Please ltd, a company registered in England and Wales under number 16665743, registered office 3rd Floor, 86-90 Paul Street, London, England, EC2A 4NE ("we", "us", "Bamya").
| Privacy contact | yavo@bamya.co |
| Data protection lead | Yana Indzheva, Director |
| UK representative (Art. 27 UK GDPR) | Not applicable — we are UK-established |
| EU representative (Art. 27 EU GDPR) | Not appointed — see below |
We are not required to appoint a Data Protection Officer and have not appointed one. We have also not appointed a representative in the EU, relying on the exemption in Article 27(2) EU GDPR: our processing of EEA residents' data is occasional, is not large-scale processing of special category data, and is unlikely to result in a risk to their rights and freedoms. Anything you would otherwise raise with a representative, raise with us at yavo@bamya.co. If our use in the EEA grows beyond that description, we will appoint one and say so here.
This policy covers both UK GDPR and EU GDPR (Regulation 2016/679).
2. The short version
Audio never reaches our servers. Your browser streams it directly to our transcription provider, in the European Union. We never receive it and could not produce a recording of your call if we were asked to. Only text comes back.
Text does reach us, and text is the whole conversation. A transcript is stored on our servers unless you turn that off, and deleted on the schedule you choose. It is also sent to our AI provider whenever you use an assistance feature — including in local-only mode, where the choice governs whether a copy stays with us, not whether one is sent.
The other people on the call did not sign up for this. They are in the transcript. Telling them, and having a lawful basis for recording them, is the responsibility of the account holder who started the capture — see §11.
3. Which hat we are wearing
| Data | Our role | Controller |
|---|---|---|
| Your account, billing, server and security logs | Controller | Us |
| Conversation content: transcripts, translations, uploaded documents, scenarios and instructions, AI write-ups, photographs | Processor | The customer whose account recorded the call |
| Lead briefs drawn from Bamya Leads | Processor | The Bamya Leads customer organisation |
Where we are a processor, our customer decides what is recorded, from whom and why; our obligations to them are in Annex A of the Terms and Conditions. If your conversation was recorded by someone else's account and you want it deleted, they are the people to ask — we will pass a request on, but we cannot lawfully act on it ourselves. See §14.
4. What we collect, and why
4.1 Your account
| What | Why |
|---|---|
| Email address, display name, hashed password, whether the address is confirmed | Running the account and signing you in. Recording is refused until the address is confirmed |
| Organisation membership and role, shadowed from Bamya Leads | Deciding which conversations you may open and which features you may use |
| Business name you registered under | Confirming the service is bought for business use |
| Plan, allowance and hours recorded per capture | Metering, billing and showing your usage |
| Billing contact and payment records | Taking payment; tax and accounting |
| Interface settings, and last sign-in time | Making the app behave as you set it; support and abuse investigation |
We keep no marketing profile, no behavioural analytics, and no advertising identifiers of any kind.
4.2 Conversation content
- Transcripts. Every line of what was said, labelled by side, with a timestamp. A translation is stored alongside the original, never instead.
- Documents you supply. CVs, adverts, background notes, scenarios and your instructions to the assistant. Uploaded
.docxand.pdffiles are read into text and the file itself is discarded — we never store the file. - AI write-ups. Minutes, debriefs, preparation notes and checklists.
- Photographs, in the visual-assistant mode only. No photograph is ever stored — not on disk, not in the database, not in any log. Each is sent to the AI provider, converted into a written description, and dropped. The description is kept; the image is not.
- Lead briefs, for accounts arriving from Bamya Leads: the contact's name and LinkedIn profile data, and the company's website text, industry, location and recent public posts. A read-only copy taken when the call is created.
- Whatever is actually said. A transcript can therefore contain health, political or religious information and other special category data (Art. 9). We do not seek it out, index it or profile on it, but we hold it whenever it is spoken. Do not use the service for calls that are predominantly such material without agreeing it with us first — see Terms §5.
4.3 Technical data
| What | Retention |
|---|---|
| Web server access logs — IP address, timestamp, page, user-agent. Query strings are stripped before writing | 14 days |
| Application error log, and application access log — IP address, timestamp, path, query strings stripped | 7 days |
| AI request log — see §5 | Same window as the transcripts it relates to |
| Failed sign-in attempts, by IP address | 15 minutes, in memory only |
| Single sign-on token identifiers | 24 hours |
5. AI features, and what is sent to the model
The assistance features work by sending your material to a third-party AI model and returning what it says.
| Feature | What goes to the model |
|---|---|
| Help | The recent transcript, your scenario, CV and instructions |
| Advice | The entire transcript, plus the same supporting documents |
| Auto-answers | The recent transcript and the pending question |
| Minutes / debrief / preparation | The entire transcript and supporting documents |
| CV tailoring | Your whole CV and the job advert |
| Word base | The CV and advert, to build a vocabulary hint for transcription |
| Visual assistant | The photograph, then its written description plus the conversation so far |
This happens regardless of your retention setting. Keeping transcripts on your own machine means we do not keep a copy — it does not mean the conversation is never sent to the model. If a conversation must never leave your machine, do not use the assistance features on it.
The AI request log. Every model request and reply is written verbatim to a restricted log so a bad answer can be diagnosed against exactly what the model saw. It therefore contains transcripts and documents, and is deleted on the same schedule as the transcripts themselves. Two things are kept out of it by design: spoken content from the transcription stream is never written to it under any circumstances, and for local-only conversations the prompts and replies are withheld entirely.
6. Who else receives your data
We do not sell personal data, we do not share it for advertising, and we have no advertising or analytics partners of any kind.
| Processor | What they get | Where | Basis for transfer |
|---|---|---|---|
| Soniox, Inc. — speech recognition and translation | Live call audio, streamed from your browser | European Union (primarily — see §7) | EU Standard Contractual Clauses and the UK International Data Transfer Addendum, incorporated into our signed data processing agreement |
| Google LLC — Vertex AI, all AI features | Transcripts, documents, photographs, prompts | European Union (primarily — see §7) | Google's Cloud Data Processing Addendum, which incorporates the EU Standard Contractual Clauses by reference in the module matching our role for that data (§3), together with the UK Addendum where the UK GDPR applies |
| Microsoft Azure — servers and backups | Everything stored: the database, backups, logs | United Kingdom | Not a restricted transfer |
| Stripe Payments UK Limited — payment processing, once billing launches | Name, email, billing address, card details (handled by them, never by us) | United Kingdom, with onward processing in the EEA and United States | A UK-to-UK transfer, so not a restricted transfer for us; Stripe's own onward transfers are covered by their data processing agreement, which incorporates the EU Standard Contractual Clauses and the UK Addendum |
Bamya Leads is a sibling product operated by the same company. Where your organisation uses both, membership information is shadowed from it and lead briefs are read out of it — an internal transfer between our own systems, not a disclosure to a third party.
We will disclose data where legally required — a court order or equivalent — and will tell the affected customer unless the law forbids it.
We do not use your conversation content to train our own models. We have none. Nor do our providers train on it: we use Vertex AI on a paid Google Cloud project, under which submitted content is not used to train Google's models, and our transcription provider is contractually bound not to use our data to train models serving other customers, and does not store audio or text by default — audio is processed transiently to produce the transcript. The paid tier is a deliberate choice: Google's free tier does permit training on submitted content, and we do not use it.
7. International transfers
- Our servers and backups are in the United Kingdom.
- Transcription is processed in the European Union. A UK-to-EEA transfer is covered by the UK's adequacy regulations for the EEA. Our agreement with the transcription provider commits them to processing within the region we have selected, while allowing limited activity outside it where necessary to operate, secure or support the service — that residual case is covered by the Standard Contractual Clauses and the UK Addendum named in §6.
- The AI features are processed in the European Union. They moved there on 9 August 2026; before that they were processed in the United States. A UK-to-EEA transfer is covered by the UK's adequacy regulations for the EEA, exactly as transcription is. Our agreement with the model provider reserves limited activity outside the region on the same footing the transcription agreement does — the region we select governs where the material is stored and where the models run, while operating, securing and supporting the service may require access from elsewhere. That residual case is covered by the Standard Contractual Clauses and the UK Addendum named in §6, and it is why that table's entry says "primarily" rather than "only".
Every routine leg is now inside the UK or the EEA. Until 9 August 2026 this section said the opposite, and said it accurately: using the assistance features sent the material in §5 to the United States whatever your retention setting said. That is no longer true, and it is the single largest change this policy has ever made in your favour.
It is not a promise of EU-only processing in the absolute, and we would rather say so than let you infer it. Providers reserve limited activity outside a selected region to operate, secure and support a service; that residual case is precisely what the Standard Contractual Clauses and the UK Addendum named in §6 are there to cover, and it is why both entries in that table say "primarily". If genuinely EU-only or UK-only processing is a hard requirement for your organisation, contact us before recording anything, and we will show you the commitments rather than summarise them.
Transfers outside the UK and the EEA are made under the mechanisms named in §6 — Standard Contractual Clauses, the UK International Data Transfer Addendum, and adequacy decisions where they apply. You may request a copy of the relevant safeguards from the contact in §1.
8. Our lawful bases
Where we are the controller:
| Purpose | Basis |
|---|---|
| Creating and running your account; providing the service; metering and billing | Performance of a contract (Art. 6(1)(b)) |
| Billing, tax and accounting records | Legal obligation (Art. 6(1)(c)) |
| Keeping the service secure — rate limiting, access logs, abuse investigation | Legitimate interests (Art. 6(1)(f)): protecting a service that holds confidential conversations |
| Detecting duplicate accounts, trial abuse and fraudulent payment | Legitimate interests (Art. 6(1)(f)): the service costs us money per hour recorded |
| Service emails about outages, changes or security | Legitimate interests (Art. 6(1)(f)), and contract |
| Marketing email, if we ever send any | Consent (Art. 6(1)(a)), withdrawable at any time |
Where we are the processor — everything in §4.2 — the lawful basis is the customer's to establish, not ours. §11 explains what that means for the person who presses record.
9. How long we keep things
Conversation content
Each account chooses where its conversations live and for how long, as a single setting with six positions:
- Local only — nothing is stored on our servers at all. The transcript stays in your browser, on that machine. See §10.
- 7 days, 15 days, 30 days, 6 months or 1 year — kept on our servers for that long and then deleted. 30 days is the default.
The choice is stamped onto each conversation as it is created and does not apply retroactively: switching to a shorter window does not delete calls already recorded under a longer one, and switching to a longer one does not extend anything already recorded.
The window runs from the last thing that happened in the conversation, not from when it was created. When it closes, a nightly job deletes the transcript and its search index entry, the minutes, debrief, preparation notes and checklist — each of those is a condensed copy of the transcript — and the corresponding entries in the AI request log. The conversation itself stays in your list, marked as expired, holding its title and dates and nothing that was said. The database is then rewritten so the deleted text is genuinely gone from the file rather than merely unlinked.
Backups. Snapshots are pruned after 7 days, but we always keep the three most recent whatever their age. So whatever window you choose, content can survive in a backup for up to roughly 7 days beyond it. Local-only conversations are never in a backup at all, having never been written to the database.
Everything else
| Data | Kept for |
|---|---|
| Account record and settings | Until the account is deleted |
| Web server access logs | 14 days |
| Application logs | 7 days |
| Billing records | 6 years (tax law) |
| Word counts per conversation | See §10 — not deleted by the sweep |
| Usage records — when a capture ran, how long, which allowance it drew on | Life of the account, then as billing records. Not deleted by the sweep: they are what an invoice is calculated from. They record that a call happened and for how long, never a word of what was said |
10. Local-only mode, and the one thing that still reaches us
You can set an account so conversations are never stored on our servers. The transcript is held in your browser (IndexedDB) on that machine and nowhere else. This is enforced by our server, not by the browser being polite: the endpoints that store lines and write-ups refuse them outright for such a conversation.
It is the least durable option, not the safest one. Clearing your browsing data deletes the transcript. The browser may evict it under storage pressure. It does not sync between devices and cannot be searched across conversations. Download anything you need to keep.
One thing derived from speech does reach us, and it is a number. For each local-only conversation we receive a count of words spoken by each side — per conversation, never per line, never the words themselves — so the charts on your home screen can include a call whose transcript we will never hold. It cannot be turned back into speech and carries no record of when anyone spoke. These counts are not deleted by the retention sweep, because totals that vanished would make every chart shrink as it aged. So a local-only conversation does leave a permanent record of how much each side said. That is a narrower promise than "nothing reaches the server", and we would rather state it than let the shorter phrase stand.
11. The other people on the call
Bamya Call records everyone in the conversation, including people who have no account with us and may not know a transcription tool is running. Nothing joins the meeting to announce itself.
If you are an account holder: establishing a lawful basis for recording the people you record is your responsibility, not ours. In practice that means telling them before the call starts and, depending on where you and they are, obtaining their agreement. Recording without consent is a criminal offence in some jurisdictions — including several US states requiring the consent of every party — and a breach of data protection law in the UK and EEA. You are the controller of that content and, under the Terms, you indemnify us for claims arising from recording someone without a proper basis. That a tool is invisible to the far end makes this easier to neglect; it does not make it lawful.
If you are the other person on a call: your words may be in a transcript held by us on behalf of the account holder who recorded it. You have all the rights in §14, but we cannot exercise most of them without that account holder's instruction, because they are the controller and we are not. Contact them first. If you cannot identify them, write to us with whatever you know — the date, the other participants, the company — and we will try to route your request. We will not disclose their identity to you unless we are lawfully required to.
12. Security
Everything is served over HTTPS. Session and sign-in cookies are HTTP-only, SameSite=Lax and Secure. The database, its backups and the AI request log are readable only by the service account. Passwords are stored as salted hashes and repeated failed sign-ins from one address are rate-limited. Every API request naming a record is checked against the caller before the route sees it, so one account cannot reach another's conversations or documents by guessing an identifier. Conversations are private by default, and sharing with colleagues is a deliberate act on one conversation, read-only, never extending to CVs. A strict Content Security Policy restricts the page to our own origin and the transcription provider. Access from a revoked organisation is cut within 15 minutes rather than lasting the life of the session.
No system is perfectly secure, and we hold no security certifications of our own.
Breach notification. If a breach affects personal data we hold as a controller and is likely to result in a risk to you, we will notify the relevant supervisory authority within 72 hours and tell affected people where the law requires it. Where we hold the data as a processor, we will notify the customer controller without undue delay so they can make their own notification.
13. Cookies and browser storage
We use no analytics, no advertising, no tracking, and no third-party cookies. There is no cookie banner because nothing here lawfully requires consent.
| Cookie | Purpose | Life |
|---|---|---|
notetaker_session | Keeps you signed in | Session / 30 days |
notetaker_remember | Signs you back in without a password | 30 days |
Both are strictly necessary, HTTP-only, and scoped to the single host that set them. Local storage on your device holds functional settings only — which microphone you chose, whether a panel is collapsed, whether you dismissed a notice — and none of it is transmitted to us. In local-only mode, IndexedDB also holds the transcripts themselves (§10). Clearing your browser's site data removes all of it, including those transcripts.
14. Your rights
Under UK and EU GDPR you have the right to access your data, to have it corrected or erased, to restrict or object to its processing, to portability, and to withdraw consent where we relied on it.
Several are available directly in the app, which is faster than writing to us: every conversation can be downloaded as a text file with its transcript and write-ups; conversations, lines, CVs and presets can be deleted outright; your name and settings are editable in Settings.
Otherwise write to yavo@bamya.co. We will respond within one month, extendable by two further months for complex requests, and we will tell you if we need the extension. There is no charge unless a request is manifestly unfounded or excessive. For deletion of an entire account, write to us — there is no self-service account deletion yet. Billing and usage records are retained afterwards where tax law requires it (§9); that covers invoices, not conversations.
A limit worth stating plainly: for conversation content we are the processor, and we must not act on an erasure or access request from a data subject without the customer controller's instruction. If you contact us about a transcript recorded by someone else's account, we will forward your request to them promptly and tell you we have done so. See §11.
15. Complaints
Please tell us first — we would rather fix it. You also have the right to complain to a supervisory authority: in the United Kingdom, the Information Commissioner's Office (ico.org.uk, 0303 123 1113); in the European Union, the authority of the country where you live, work or where the issue arose, listed by the European Data Protection Board at edpb.europa.eu. Complaining does not affect any other remedy you have.
16. Automated decision-making
The service generates suggestions — what to say next, a summary, a tailored CV. None of it is a decision. No output produces a legal or similarly significant effect, and every one is read by a person who decides whether to act on it. We do not carry out automated decision-making or profiling within the meaning of Article 22.
Two points on accuracy, which matter most to people who are in a transcript rather than using the app. Transcription is imperfect — words are misheard, especially across accents, names and technical terms, so a transcript is a working record and not evidence of what someone said. And speaker attribution can be wrong: where both voices share one microphone they are separated by an algorithm that is explicitly unreliable, so a line attributed to one person may have been said by another.
17. Children
Bamya Call is not intended for and is not offered to children. Accounts are for adults acting in the course of a business, profession or trade. If a child's voice appears in a transcript, the account holder who recorded it is the controller and bears the corresponding responsibility.
18. Changes to this policy
We may update this policy. Where a change materially affects how we handle your data — a new processor, a new purpose, a shorter retention default — we will tell account holders by email before it takes effect, and will not apply it retroactively to data already collected where that would reduce its protection. The version and date at the top of this document are authoritative.
19. Contact
ATTN Please ltd
3rd Floor, 86-90 Paul Street, London, England, EC2A 4NE
yavo@bamya.co
See also the Terms and Conditions, including Annex A, which contains the data processing terms that apply where we act as your processor.