Privacy Policy
This policy explains how MLOG collects, uses, stores and deletes your data. The one-line version: your meeting data is used only to generate your briefs, answers and minutes — never to train models, never sold.
01Core promises
These promises are built into the product’s behavior — each one you can point to in the interface:
- Your data is used only to generate your briefs, answers and minutes. We use your meeting content for no other purpose.
- Never used to train models. Recordings, transcripts, minutes and your knowledge base never enter model training — not ours, and not our vendors’, who are bound by the API terms under which we use them.
- Local-first. Audio is AES-256 encrypted onto your own device before any upload — on all four surfaces, including the web app and the browser extension (into the browser's private storage, OPFS); capture starts only when you press “Start recording” — no hidden capture, no background listening.
- We never sell personal information. No selling, renting or trading; no sentiment analysis, no mood scoring.
02Data we process
- Account information: name, email address, company and workspace settings — for sign-in, collaboration and billing.
- Meeting content: the audio you actively record, transcripts, minutes and linked documents — used only to generate your outputs and power your search. Titles, attendee details, agenda drafts and selected sources that you submit for a pre-meeting preview, and the candidate content generated from them, are also used to prepare your meeting.
- Usage and device data: crash reports, feature usage and credit metering — only for reliability, security and billing, itemised and auditable. Crash reports and usage events are processed by PostHog in the EU; they carry a pseudonymous account identifier and never carry meeting content. You can switch them off for this browser right here on this page, and signed-in users can switch them off for the whole account under Settings → Data & privacy.
- Cookies and local storage: we set no advertising or tracking cookies. Here is everything the website and the web app keep in your browser, and why: a sign-in token (clearing it signs you out); your interface language, your microphone choice and whether you have finished onboarding; for analytics, a pseudonymous device identifier and your on/off choice, so the same browser is not counted twice; for recording, a local encryption key (kept until you clear this site's data) and the audio segments that have not uploaded yet, each deleted as soon as it has uploaded; and, for the current tab only, a check that ties a sign-in with an outside account to this browser, a one-time notice after a workspace has been deleted, and a fixed marker indicating a web visit from the Mac App Store app. This marker only adjusts what the current tab displays and contains no identity data, timestamps or meeting content. For an account closure that has not finished, we also keep the email you confirmed, basic account details to restore the closure screen, and a marker that checks whether the browser session still matches. The app tries to clear this record when it learns that closure is complete or that the server explicitly rejected the request, when you sign out or successfully sign in again, or when recovery detects a session mismatch. Browser storage restrictions can prevent removal; the record may remain until you clear this site's data. The rest stays until you clear this site's data or, for analytics, switch analytics off above. Apart from the analytics events described above, the recorded audio as it uploads, the fixed source marker in a visit link sent to our website with the page request, the sign-in token sent to our own servers with each request, and the confirmed email sent to our own servers through the original signed-in session to request or retry the same account closure, none of it leaves your browser. The local recovery record cannot authorize an account closure on its own.
This choice is remembered in this browser only, and it takes effect immediately — no account needed. If you are signed in, the switch under Settings → Data & privacy applies to your whole account, on every device.
03Storage & transmission
From your microphone to our database, data takes exactly three steps — encrypted at every one:
- Local: on the desktop and mobile apps, audio is AES-256 encrypted onto your own disk before anything is uploaded — offline-safe, with at most 5 seconds lost in a crash. The web app and the browser extension write each segment to this device first as well, AES-256 encrypted into the browser's own private storage (OPFS). The download page sets out how the surfaces differ. How long a local copy stays depends on the app: the desktop app keeps your original recordings on your own device up to the on-device cache limit you set, and rotates out the oldest already-uploaded segments only once that limit is reached; the mobile app, the web app and the browser extension delete each local segment as soon as it has uploaded.
- In transit: uploads happen only when you are online and have authorized them; TLS 1.2+ end to end, no plaintext along the way.
- Cloud: encrypted at rest — the database and the object store both in the Asia-Pacific region, with no cross-region replication configured. Workspaces are strictly isolated and meeting content is workspace-only by default. Where your data is *processed* is a different question; “Sub-processors” below names every party that touches it.
04Retention & your right to delete
- Cloud audio is kept for 90 days by default, and 365 on the two higher paid tiers, counted from when each segment is written rather than from the end of the meeting; every Monday you get one digest listing what expires in the next 14 days. Transcripts, minutes and quoted evidence have no expiry — they stay until you delete the meeting or close the account, whether or not you are still subscribed. Local copies are yours, and cloud expiry never touches them.
- Inputs and results for pre-meeting agenda previews, and their links to individuals or meetings, are kept for 24 hours. Access ends then, and the content and links are removed within the following hour. Workspace usage records without content or personal links are kept until 35 days after that billing period ends. Account closure, removal from a workspace and verified attendee erasure requests clear the affected preview content and personal links. Deleting the workspace also removes these usage records. Formal briefings follow the meeting-output retention rules above.
- You can export or delete all of your data anytime under Settings → Data & privacy.
- Deletion cascades: audio, transcript, minutes, evidence links, search index and share links are removed together, and a confirmation is sent to you. Encrypted database backups are held for a limited period in the same region and are overwritten on a rolling basis; data in a backup is never restored into production except for disaster recovery.
- Closing your account deletes your identity, sessions, notifications, avatar, consent records, every calendar or integration you connected in any workspace and the meetings you kept private and shared with no one — and, for every workspace where you were the last remaining member, its meetings, recordings and knowledge base. Two things stay on purpose: the usage ledger and the billing records, which tax and accounting law require us to keep and which are unlinked from your identity; and the audit entry recording that the erasure happened, which is the only proof left that it did. In a workspace that other members still use: the meetings you shared with the workspace or with a colleague, and the workspace's knowledge base, stay for the team; where you appear as a meeting participant, the entry is kept so that what was said there stays attributed, but your name and email are removed from it; and the workspace keeps its name, even one that contains yours (such as “Alex’s workspace”), because the remaining members rely on it — our legitimate interest under GDPR Art. 6(1)(f) — and its owners and admins can rename it at any time. Support requests you sent are deleted with your account.
05Speaker identification
MLOG creates no voiceprints, stores none and matches none, and it processes no biometric data. Speakers are separated by the transcription model; you name them, and unnamed ones stay as “Speaker N”. Any label can be corrected by hand, and corrections apply to that meeting only.
06Your rights (GDPR / CCPA)
Wherever you are, we honor your rights to the GDPR and CCPA standard:
- Access, rectification, export (portability), erasure, restriction and objection — initiate them in-app under Settings → Data & privacy, answered within 30 days.
- For California residents: the rights to know, delete and opt out of sale equally apply — in fact, we sell no personal information at all.
- You can also exercise any of these rights by writing to support@mlog.ai.
07Sub-processors
We engage the following categories of sub-processors on a least-privilege basis. Write to us and we will send you the current named list.
| Category | Purpose | Residency | Constraints |
|---|---|---|---|
| Cloud infrastructure | Hosting, storage and backup | Singapore (database) · APAC (object storage) | Encrypted at rest · no cross-region replication |
| ASR providers | Speech-to-text (several, hot-swappable) | United States · global | Not used for training (bound by vendor terms) |
| LLM providers | Minutes generation and advisor Q&A (several, hot-swappable) | United States · global | Not used for training (bound by vendor terms) |
| Search-indexing providers | Search indexing of transcripts and knowledge-base text, and ranking of search results | United States | Not used for training (bound by vendor terms) |
| Document recognition (OCR) | Reading PDFs and images uploaded to the knowledge base, and scanned pages the desktop app could not read on its own | United States | Not used for training (bound by vendor terms) |
Sub-processors are listed by category so the list survives a vendor hot-swap. Five kinds are not in the table because they never touch meeting content, and we name them here for completeness: payments (receives your email address and the last four digits of your card), transactional email and push delivery (receive your address or device token and the notification text, meeting title included), the hosted job runner that schedules post-meeting processing (receives identifiers only), and — since 21 August 2026, when in-app purchases were switched on — the subscription service behind the App Store and Google Play (receives a workspace identifier and purchase events, never any meeting content).
08Data Processing Agreement (DPA)
Where your organisation acts as controller and requires a written data processing agreement, or Standard Contractual Clauses for transfers out of the EEA, write to support@mlog.ai and we will arrange the applicable terms with you.
09Browser extension
The MLOG browser extension does one job: it records a meeting — the audio of the tab you choose and, if you allow it, your own microphone. It has no content script, so it never reads the contents of any page you visit; the only thing it reads from a tab is its title. It records nothing until you press “Start recording”. Here is what each permission it asks for is actually used for:
- Tab audio capture: to record the audio of the tab you are on, from the moment you press “Start recording” until you press stop. Video is never requested — not a single frame of your screen is captured.
- Microphone: to record your own voice, through the browser's usual microphone prompt and only while a recording is running. If you decline, only the tab is recorded; if you allow it later, your microphone joins the recording from that moment.
- Background document: to keep the recording loop running. The extension’s popup closes the moment it loses focus, so the recording cannot live there.
- Local storage: to remember that a recording is in progress, so one the browser cut short can still be uploaded afterwards. It holds a meeting identifier, a timestamp and the tab's title — no audio and no transcript. Audio that has not uploaded yet waits in the storage the browser keeps for this extension and is deleted as soon as each segment has uploaded.
- Sign-in: to connect the extension to your MLOG account once, without typing your password again.
- Access to our API only: to create the meeting and upload the recorded audio. The extension talks to no other host.
- Tab title: read when you open the extension, so it can show which tab it is about to record, and sent to our API once, when you press “Start recording”, as the meeting's name.
MLOG’s use and transfer of any information received through the browser extension adheres to the Chrome Web Store User Data Policy, including its Limited Use requirements. Extension data is used only to provide the features described here — never sold, never used for advertising, never used to train models.
What the extension promises about your recording is not the same as the desktop app, and the difference is on the download page — please read it there before you rely on either.
10Recorded in a meeting, without an MLOG account
You may be in MLOG without ever having used it, because someone you met with recorded the meeting. This section is for you. It covers what the law requires when data about a person comes from someone else (GDPR Art. 14).
- What is held, and where it came from: the recording of the meeting, a verbatim transcript of what you said with timestamps, the speaker label you appear under, your name and email address if the person recording entered them, and the minutes, action items and quotes generated from all of that. None of it came from you: the organisation that recorded the meeting gave it to us to process.
- Who is responsible for it: the organisation that recorded is the controller: it decides whether to record, how long to keep the record and who sees it. We are its processor. Telling you about the recording is its duty, not ours, and we send meeting participants no notice of our own.
- The notice page: the recording disclosure MLOG gives the person recording ends with a short link — mlog.ai/recorded/ followed by a six-character code. If they read it out or pasted it into the meeting, that page names the organisation that recorded, the privacy contact and legal basis it has given (or says it has given none), how long it keeps the audio, and what you can ask for. The page does not track you: we count how many times it is opened each day, and briefly count visits per network address so that the codes cannot be guessed in bulk.
- Making a request: on that page you can ask to see what the record holds about you, to have the parts about you deleted, to object to further processing, or to correct how you are labelled as a speaker. We check only that the email address you give is yours, by sending it a six-digit code — no identity document, no account. The request goes to the owners and admins of the workspace that made the recording, with a 30-day deadline, and they answer you directly. We do not decide it for them.
- No code, or no answer: you can also make a request at mlog.ai/recorded without a code: say who recorded and roughly when, and our team passes it to that workspace. We will not tell you whether a particular organisation uses MLOG. If the deadline passes without an answer, write to the support address in the receipt we emailed you, quoting your reference, and we will tell you who the controller is and how to complain to a data protection authority.
- What we keep about your request: your email address, the type of request, the meeting details and note you gave, and when and how it was answered. We keep this as the record that the request was handled, which the law requires of us (GDPR Art. 6(1)(c)); it therefore stays even if the organisation later deletes its workspace, and we use it for nothing else. To stop the form being abused, requests are also counted per network address for a short time; that count is not stored with your request.
- What we do not do with it: we do not market to you, we do not build a profile of you across meetings or organisations, and we do not sell or trade any of it.
11Changes & contact
Material changes to this policy are announced 30 days in advance by email and in-app notice. For any privacy question, write to support@mlog.ai — whatever your plan, we confirm receipt within one business day and answer requests about your personal data within 30 days.