You took part in a meeting that was minuted with Protoki

The controller within the meaning of the GDPR is the organisation that issued the invitation — that is, the organisation using Protoki. It decides whether a recording is made, how long the data is kept and who can see it. Address your requests to that organisation first; we will forward them on request (section 8).

rooom AG operates Protoki on behalf of that organisation as a processor under Art. 28 GDPR and pursues no purposes of its own with your data. Where rooom AG issued the invitation itself, it is also the controller.

rooom AG, Löbstedter Str. 47a, 07749 Jena, Deutschland · info@rooom.com
Data protection officer: Landgraf Datenschutz GmbH, Markt 22, 07743 Jena · dsb@landgraf-datenschutz.de

Protoki joins online meetings (alternatively, audio files are uploaded), the recording is transcribed and AI-assisted minutes with tasks are created from it. The purposes are:

• Business documentation: minutes, overviews and tasks from meetings.
• Assigning meetings to client contexts based on the email domain of external participants.
• Only if the inviting organisation has switched on live translation for the meeting: live captions during the meeting and their machine translation into the other languages of the meeting, so that participants who speak different languages can follow each other. If the inviting organisation has also switched on “Read aloud in headphones”, anyone reading the live captions in Protoki, via the guest link or in the Protoki side panel in Microsoft Teams can have the translation read out in their language by a synthetic voice.
• Translation of finished minutes into further languages — at the request of a person who may see the meeting, or automatically into the languages stated for the meeting.

No performance or behaviour monitoring takes place. No speaking-time statistics and no personal evaluations across meetings are produced, and there is no emotion or sentiment analysis. No permanent voiceprints are stored; speaker-separation data exists only temporarily while a meeting is being processed and is deleted together with the audio recording.

• Audio recording of the meeting (temporary only, see section 6).
• Transcript with speaker assignment.
• Public chat messages from the meeting (for example shared links), where the platform makes them available to the recording participant — they are included in the minutes.
• Technical meeting metadata from the recording service (the platform's participant list, start and end of the recording) — the basis for the attendance details in the minutes.
• AI-generated minutes, including the tasks assigned in them.
• Participant metadata: display name from the meeting platform, where applicable the email address from the calendar invitation, approximate time present.
• With live translation switched on: the sentences recognised during the meeting, with time, display name from the meeting platform and recognised language, and their machine translations (live captions) — in substance the same as the transcript, only earlier and in further languages. If a person has the translation read out (“Read aloud in headphones”), a synthetic speech output is also produced from the translated sentence — with one fixed voice, not the voice of the person speaking.
• Machine-translated versions of the minutes, where they were requested or created automatically.

Source (Art. 14(2)(f) GDPR): from the meeting itself (audio, the platform's participant list), from the corresponding calendar invitation of the inviting organisation and — where the speaker assignment is corrected afterwards — from manual entries made by users. The recording is announced before it starts: by the bot name with the addition “is recording”, by a notice in the meeting and, where applicable, by an email sent before the meeting.

• External participants: as a rule Art. 6(1)(f) GDPR — legitimate interest in documenting business meetings, based on a documented balancing of interests. You have a right to object (section 7).
• Employees of the inviting organisation: Art. 6(1)(b) and (f) GDPR, as a rule supplemented by an internal usage policy or a works agreement that defines permitted meeting types, access rights and retention periods and rules out performance monitoring.

Which basis applies in a specific case is decided by the inviting organisation as the controller.

To operate Protoki, rooom AG uses the following service providers as sub-processors (Art. 28 GDPR). Processing and storage take place in the European Union; we name the “group headquarters” regardless, because some of these providers are companies headquartered outside the EU.

• Spoke SAS („Meeting BaaS") — The recording bot joining the online meeting, and the audio recording; with live translation switched on, also forwarding to real-time speech recognition. Processing location France (EU), group headquarters France.On our deletion request the provider removes the recording, transcript, chat history and the names of participants and speakers. The session metadata — bot ID, status, timestamps, duration and token consumption — remain with them, and that includes the meeting URL. According to the provider on 31 August 2026 these rows are currently retained indefinitely; deletion or redaction is possible only on request via their support.
• Scaleway SAS — engaged through Meeting BaaS — The data centre and storage on which Meeting BaaS runs its service. Processing location Paris, France (EU), group headquarters France.
• Gladia SAS — engaged through Meeting BaaS — Real-time speech recognition within the Meeting BaaS offering — only with live translation switched on. Processing location France (EU), eu-west region, group headquarters France.Protoki switches off transcription at Meeting BaaS and transcribes itself after the meeting. On 31 August 2026 Meeting BaaS confirmed in writing that with transcription disabled no audio is sent to Gladia. Since 15 October 2026 there is exactly one exception: only for meetings with live translation switched on does the audio go via Meeting BaaS to Gladia for real-time speech recognition, in the eu-west region (France). For all other meetings the confirmation continues to hold as long as the provider’s own transcription stays switched off. For this exception it has not yet been confirmed in writing whether Gladia uses the audio and transcript for training, how long it retains them, and whether our deletion request to Meeting BaaS reaches them there; we are currently obtaining this information.
• Mistral AI SAS — Transcription with speaker separation, translation of the live transcript, production of minutes and tasks, and translation of finished minutes. Processing location France (EU), group headquarters France.
• Supabase, Inc. — Database and file storage of the application. Processing location Frankfurt am Main, Germany (EU), group headquarters United States.
• Vercel, Inc. — Operation and delivery of the web application. Processing location EU region (Frankfurt am Main), group headquarters United States.
• Plus Five Five, Inc. („Resend") — Sending of the minutes emails and system emails. Processing location EU region (Ireland) for sending, group headquarters United States.
• rooom AG (SovrGPT) — Speech output for the live translation (“read aloud”); intended as a second model provider for minutes and tasks. Processing location Germany (EU), group headquarters Germany.Only for meetings with live translation and only if the company has switched on “read aloud in headphones”. One fixed synthetic voice, no voice cloning. Not currently used for minutes.

“Engaged through” means: this provider has no contract with us, but with the service provider named before it (Art. 28(4) GDPR). We name it regardless, because it sees the data.

In addition, those people at the inviting organisation who are entitled to the minutes are given access. No data is passed on for advertising purposes.

With live translation switched on, additionally: participants of the meeting, including those without a Protoki account, who read the live captions via the guest link or the Protoki side panel in Microsoft Teams. The link is shared in the meeting, usually at the start in the meeting chat, and works only for this meeting. Anyone who has the link can read along without having been admitted to the meeting — so please do not pass it on. If the inviting person sets it up, the live captions also go to Microsoft Teams as one shared caption line (CART), back into the same meeting. Real-time speech recognition is handled by Gladia via Meeting BaaS, translation by Mistral AI. If the inviting organisation has switched on “Read aloud in headphones”, rooom AG (SovrGPT, based in Germany) produces the speech output; for this SovrGPT receives only the translated sentence as text, neither audio nor display names (see the list above).

• Audio recordings: deleted after successful transcription. The inviting organisation can set a limited retention period — at most 12 months; after that the recording is deleted automatically. Uploaded files without a speaker assignment are deleted after 7 days at the latest. If processing fails or no speech is detected, the recording is kept for another attempt for at most 30 days (longer only if the inviting organisation has set a longer retention period, at most 12 months) and is then deleted automatically. For meetings with live translation there are two exceptions: if the transcript comes from the live recognition, the recording is kept for at least 7 days (longer only if the inviting organisation has set a longer retention period, at most 12 months) so that the meeting can be reprocessed if the live recognition had gaps. If no transcript is produced at all, because the subsequent speech recognition cannot handle one of the stated languages and the live recognition delivered nothing, the recording remains stored so that the inviting organisation can resolve the matter — for a maximum period of 30 days (longer only if the inviting organisation has set a longer retention period, at most 12 months); it is then deleted automatically.
• Transcripts: 12 months by default.
• Minutes: 24 months by default.
• At the bot provider: recording, transcript, chat history as well as participant and speaker names are deleted as soon as Protoki has taken over the recording. The bot's session metadata — identifier, status, timestamps, duration, token consumption and the meeting URL — remain stored at the provider; according to the provider on 31 August 2026 indefinitely. Redaction of the meeting URL or deletion of these details is possible only on request via the provider’s support. With live translation switched on, it has not yet been confirmed in writing how long Gladia, the real-time speech recognition within this provider’s offering, retains audio and transcript, and whether the deletion request reaches them there; we are currently obtaining this information.
• Live captions and translations: live captions are deleted together with the transcript, at the latest 12 months after the respective sentence; translated minutes together with the minutes. The guest link itself is not stored; it works only while live translation is switched on for the meeting, and at most until two hours after the meeting has ended. Protoki stores the speech output of “Read aloud in headphones” neither in the database nor in file storage: it is produced only for sentences at most 5 minutes old, is held only transiently in the server’s memory until newer sentences displace it or the server restarts. The listening person’s browser may reuse it for up to one hour without a new request; how long it keeps it after that is up to the browser. The sentences sent to SovrGPT are the stored translations of the live captions and are deleted at Protoki together with them. We do not yet have a written statement from SovrGPT on how long the sentences sent to it are retained there.

Database backups expire after 7 days at the latest; after a restore from a backup, deletions that have already been documented are carried out again.

You can object to a recording during the meeting — verbally or in the meeting chat. The recording bot is then removed.

You have the right of access (Art. 15), to rectification (Art. 16), to erasure (Art. 17), to restriction of processing (Art. 18) and to data portability (Art. 20 GDPR). These rights are directed against the responsible organisation. If you do not know which one that is, write to us: we will name the body or forward your request — you need neither an account nor any form other than the one above.

We confirm receipt and reply within one month (Art. 12(3) GDPR). Where your data is collected from a calendar invitation, the responsible organisation will inform you within one month at the latest, and earlier on first contact (Art. 14(3) GDPR).

You may also lodge a complaint with a data protection supervisory authority (Art. 77 GDPR) — for rooom AG this is Thüringer Landesbeauftragter für den Datenschutz und die Informationsfreiheit, Häßlerstraße 8, 99096 Erfurt.

• No voiceprints and no recognition of voices across meetings.
• No emotion, sentiment or engagement analysis.
• No speaking-time or performance evaluation.
• No use of your content to train AI models — neither by us nor by our providers. For one provider we cannot yet substantiate this: with live translation switched on, Gladia (France, EU) processes the audio in real time via Meeting BaaS, and we are currently obtaining written commitments on training and deletion for this route.
• No chat with the bot; it accepts no input from the meeting.

Minutes are created with the help of AI and are labelled as such. The verbatim transcript and the generated minutes remain separate. The same applies to machine translations: live captions in translation and translated minutes are labelled as AI-generated in Protoki, via the guest link and in the Teams side panel; the original is kept. The shared caption line in Microsoft Teams (CART) carries no label of its own on each line. The speech output of “Read aloud in headphones” is a synthetic voice; while “Read aloud” is switched on, “Synthetic voice · AI translation” is shown next to it. No participant’s voice is imitated. AI results can be wrong.