Live captions and translations built for PIPEDA, Quebec Law 25 and GDPR obligations
Transcripts are personal and often confidential. InterScribe keeps them encrypted, keeps them only for a set window, erases them on request with a record, and shows who did what.
Encrypted
In transit and at rest, every store
Bounded
A retention window for each kind of data
Recorded
Every erasure, with what was removed
Quick summary
What it is
Transcripts are personal and often confidential. InterScribe keeps them encrypted, keeps them only for a set window, erases them on request with a record, and shows who did what.
Who it's for
Designed for event teams, education programs, ministries, and enterprises running multilingual live or recorded communication.
Proof
Encrypted In transit and at rest, every store | Bounded A retention window for each kind of data. Setup in minutes with attendee access via web link or QR and no app downloads.
Procurement, privacy officers and attendees each ask a different question
A security questionnaire wants isolation, retention and audit evidence. A privacy officer wants to know where transcripts go, for how long, and how a deletion request is fulfilled. An attendee wants to know their words are not kept forever. The same architecture has to answer all three.
- Transcripts copied into tools that never delete them.
- No way to prove that a session or an organization's data was actually erased.
- One retention period for every customer, whatever their contract says.
- Recordings readable by anyone who guesses a link.
- No trail of who corrected, exported or removed a caption.
Privacy controls that are part of the pipeline, not a policy on top of it
Every caption, translation and correction carries the organization it belongs to. Retention, erasure, access and audit are enforced where the data lives, and the evidence is available to your privacy officer without asking engineering.
STEP 1
Know how long transcripts live
Defaults: the live stream buffer is kept for 24 hours after a session ends, word-level segments for 90 days and final captions for 2 years. A retention job applies these windows every hour. Ask us for different windows for your organization and we set them.
STEP 2
Erase on request, with a record
Deleting a session or an organization erases its captions, translations, corrections and evaluation copies everywhere they exist, and writes a completion record showing what was removed and when.
STEP 3
Show your auditors the trail
Every privileged action in the operations console and every stream connection is logged with who, what and when, never with the transcript text.
Key capabilities
Encryption everywhere
TLS in transit; encryption at rest for the database, the in-memory stream store and the object store.
Organization isolation
Personal recordings are readable only by their owner and their organization's staff; private events only by members or an event-scoped access token.
Retention windows
Windows for captions and segments, applied automatically every hour. Ask for shorter windows in your contract and we set them for your organization.
Verified erasure
One operation removes a session or an organization from every store, including evaluation copies, and records the result.
Minimal data
The caption pipeline does not store audio; recordings exist only for sessions that are recorded. Speakers are numbered, never named. Application logs never contain transcript text.
Known processors, your choice
Translation providers are listed and can be limited per organization, for example to a single provider.
Frequently asked questions
Talk to a personAre you certified compliant with PIPEDA, Law 25 or GDPR?
These laws do not certify products. InterScribe is built to support the obligations they place on you as the organization collecting the data: minimization, bounded retention, erasure, safeguards, access controls and records. We provide the documentation your privacy officer needs to assess it, including a data inventory and a subprocessor list.
Where is data processed?
Our database, cache and object storage run in Amazon Web Services in the United States (Ohio, us-east-2). Speech recognition and translation providers process audio or text under their own terms; the subprocessor list names them. PIPEDA and Law 25 allow this when attendees are informed; our privacy policy and your notice to attendees should say so. If your organization requires Canadian residency, tell us before you sign and we will answer in writing.
Which third parties see transcript text?
Speech recognition receives audio; translation providers receive the source text of a caption and a few previous captions for context. They process it under their own terms, and Deepgram, our speech recognition provider, may use audio to improve its models unless that is switched off. The full list, and what each receives, is in our subprocessor list, available on request and with a data processing agreement.
How do I fulfil a deletion request?
Delete the session or the organization in the Portal, or ask us. The pipeline erases every copy, including evaluation exports, and produces a record of what was removed for your files.
How long do you keep captions by default?
The live stream buffer for 24 hours after a session ends, word-level segments for 90 days, final captions and corrections for two years. Ask us to shorten or lengthen these windows for your organization.
Does this page replace legal advice?
No. It describes what the product does. Your obligations depend on your jurisdiction, your attendees and your contracts; validate them with counsel. We are happy to answer their questions.
How does Ask (viewer questions) handle privacy?
Ask lets viewers ask questions about a session; answers are generated by AI from the session's transcript. Question text, the transcript and the answer are processed by our AI providers (OpenAI, and Anthropic and Cohere through Amazon Bedrock) to produce answers and search past sessions. Your question and its answer are stored in our database and retained with the session. Where the organizer has turned on search across past sessions, excerpts of session transcripts are also kept in an Amazon S3 Vectors index and removed when the session is deleted. Organizers see the questions and answers, but never who asked. The asker's account is kept only to show them their own questions and apply rate limits, and is never shown to organizers.