Meeting transcription for legal and advisory teams — kept inside your firm
Not a certified transcript
What Khulasa produces is an AI transcript for internal use: the full text of the meeting as spoken, with speaker labels and timestamps, plus a summary, the decisions and the action items. It is not a certified court record, it is not prepared by a sworn or licensed transcriber, and it is not a substitute for one where a court, a regulator or a notary requires it. Treat it the way you would treat a good associate's notes: fast, searchable and easy to check — click any line to jump to that moment in the audio — and reviewed by a lawyer before anything is relied on. There is no live transcription during the call: everything is produced after it ends. Khulasa gives no legal advice; how a recording may be used in your jurisdiction is a question for the firm.
Client calls in Arabic and English
Client work in the region rarely stays in one language. A client explains the facts in Arabic, counsel replies with the English term of art, a contract clause is read out in English and discussed in Arabic. Khulasa transcribes Arabic and English in the same meeting — even inside the same sentence — and keeps English names, tools and terms in Latin script instead of rewriting them in Arabic letters. Choose Auto-detect, Arabic + English, English or Arabic when you request the recording; the summary, decisions and action items come back in the language that dominated the conversation. Speaker labels and timestamps run through the transcript in both languages, so counsel can go back to exactly what the client said. We make no claim for specific dialects or for other languages, and the app itself is fully bilingual with a right-to-left Arabic interface.
Nothing trains a model unless you opt in
Your meeting content is processed to produce your analysis and to run and secure the service. We do not use it to train AI models — training is opt-in, and it is off by default. Recordings, transcripts and summaries are encrypted in transit and at rest and isolated per organisation: every organisation's data is strictly separated, and your matters never mix with anyone else's. Media is never handed out through a public storage URL — it is streamed through the API behind your session. The sub-processors we rely on are listed publicly, and a data processing addendum is available to every firm.
Inside the firm: on-premise or private cloud
A firm that cannot send privileged audio to a cloud provider does not have to. Private Cloud is a dedicated, isolated deployment we run for you, hosted where you need it — the location is agreed with you when we set it up. On-Premise is the same system installed on your own hardware, able to run with no outbound connection at all. Both replace the managed cloud models with self-hosted speech and language models, so your audio is never sent to a third-party AI provider, and both are quoted per organisation. The same bot joins Google Meet, Zoom and Microsoft Teams from inside that deployment, so the firm's way of working does not change.
Consent and disclosure
Khulasa never records silently, which matters more in a client call than anywhere else. It joins as one extra participant, muted and with no camera, under a display name your organisation sets with “(Recording)” appended, so it is visible in the participant list. As soon as it is admitted, and before it starts recording, it posts a notice in the meeting chat in English and then in Arabic, saying that the meeting is being recorded to produce a summary and linking to the published recording notice, which explains what is captured and how a participant can object. The notice is posted under the same display name the bot joined with, so the participant list and the chat never show two different names. Anyone who would rather not be recorded can ask the host to remove the bot, and the recording ends the moment it is removed. Khulasa does not use the recording feature built into Google Meet, Zoom or Microsoft Teams, so their native recording indicators are not shown; the named participant and the chat notice are the disclosure. We recommend that counsel also mentions the recording when the call opens, and that you check your bar's rules and local law on recording conversations before you start — consent requirements differ by jurisdiction, and that judgement is the firm's.
Read the recording noticeHow the bot joins and what participants see
Redaction, retention and legal hold
Khulasa can redact email addresses and phone numbers before the analysis is stored — those two categories, not names, not case numbers, not anything else — so a transcript that circulates inside the firm carries fewer contact details than the call did. You can set a retention window after which recordings are erased everywhere, and place a legal hold on a single recording to exempt it from that window while a matter is open. Anything can be deleted permanently at any time by the people you allow; deleted media is erased immediately, and copies of the associated database records may persist in encrypted backups for up to fourteen days after deletion.
- Redaction — email addresses and phone numbers, before the analysis is stored
- Retention window — recordings erased everywhere once it passes
- Legal hold — exempts one recording from the window while the matter is open
- Permanent deletion — at any time; database copies may persist in encrypted backups for up to fourteen days
Access control and audit
Who may see a client call is decided by the firm, not by whoever has the link. Every recording belongs to your organisation and is visible according to the rules you set: private to its owner, shared with a group such as a practice team, or open to the whole workspace, plus explicit per-person shares. Organisation roles, workspace groups and per-recording sharing decide exactly who can see what, and colleagues who are not on the list can request access rather than be handed a copy. Public share links stay off until an admin enables them, and expire by default. A tamper-evident audit log records who did what, and can be exported as CSV when a client or a regulator asks. Firms with their own identity provider can sign in through it (OIDC) on the Private Cloud and On-Premise plans. Any member of the workspace who is allowed to record can request a recording, while roles and permissions decide who may view, share and delete.
Works where your clients already meet
Clients pick the platform, so the notetaker has to follow. Khulasa joins Google Meet from its own Google account, Zoom as a browser guest with the invite link, and Microsoft Teams as an anonymous guest from the lobby — nothing is installed in the client's tenant, and no Copilot or Zoom licence is needed. Any member of your workspace who is allowed to record can request a recording of a meeting they were invited to, as long as the host admits the bot. If nobody admits it within 5 minutes, the recording fails without capturing or billing anything; if the bot is removed, the recording ends and it does not rejoin on its own.
Google Meet AI notetakerZoom AI notetakerMicrosoft Teams AI notetaker
Questions legal teams ask
Is the transcript a certified court record?
No. It is an AI transcript with speaker labels and timestamps for internal use, and it is not a certified court record; it is not prepared by a sworn or licensed transcriber and does not replace one where a court, a regulator or a notary requires it. Review it before relying on it. Khulasa gives no legal advice.
Is our client data used to train AI models?
No. Your meeting content is processed to produce your analysis and to run and secure the service; we do not use it to train AI models. Training is opt-in, and it is off by default.
Can recordings stay inside the firm's own infrastructure?
Yes. On-Premise installs the whole system on your own hardware and can run with no outbound connection at all; Private Cloud is a dedicated, isolated deployment we run for you, hosted where you need it — the location is agreed with you when we set it up. Both use self-hosted speech and language models, so audio is never sent to a third-party AI provider, and both are quoted per organisation.
How is attendee consent handled?
Openly. The bot appears in the participant list under a name your organisation sets with “(Recording)” appended, and posts a bilingual notice in the meeting chat as soon as it is admitted, linking to the recording notice that explains what is captured and how to object. Anyone can ask the host to remove it, and the recording ends the moment it is removed. Whether that satisfies the consent rules of your jurisdiction is for the firm to decide; we recommend that counsel also announces the recording.
Can personal data be redacted before it is stored?
Email addresses and phone numbers can be redacted before the analysis is stored. That is the extent of it — names, case numbers and other identifiers are not redacted — so treat the transcript as confidential firm material and rely on access control, retention and deletion for the rest.
Does it handle Arabic and English in one client call?
Yes. Khulasa transcribes Arabic and English in the same meeting, including sentences that switch between the two, and keeps English terms in Latin script. The summary and action items come back in the language that dominated the call, and the transcript is produced after the call ends, not live. We make no claim for specific dialects or for other languages.