No — Sageio does not train AI models on your meeting content, and the AI vendors we send audio to for transcription and translation are contractually restricted from doing so either. That second half is the part that matters, and it's the part this page exists to explain.
Anyone can write "we don't train on your data" on a marketing page. We'd rather show you where the promise actually lives — in the contracts, the retention table, and a public list of every company that touches your data — so you can verify it instead of trusting it.
The promise, precisely
Here is the claim, stated the way it appears in our binding documents rather than the soft way:
- Sageio does not use your meeting content to train AI models. Our Data Processing Agreement lists meeting audio with the retention line "processed in memory only; never written to storage. Not used for model training." That's a contractual term, not a blog sentence.
- Our AI vendors are barred from training on it too. Your audio passes through specialist speech-to-text and translation providers. Each one processes your content under documented no-training terms — the arrangement most meeting tools can't show you, because they've never put their vendor chain in writing.
- We create no voice profiles. The DPA states it plainly: Sageio does not use meeting content to identify any person by voice and creates no biometric templates.
If a promise isn't in a document you can hold us to, treat it as marketing. These three are.
What actually happens to your audio
During a meeting, audio streams to our speech and translation vendors, gets transcribed and translated in memory, and is discarded. It is never written to storage — not by us, not as a vendor training set, not "temporarily."
What's kept is text: the transcript, translations, and the AI summary. That text is encrypted in transit and at rest, stored in the region you choose (US, EU, or APAC), and retained until you delete it or close your account. Retention is your call, not ours.
When you leave, the exit is defined too. On termination, we delete or return all customer personal data within 30 days at your choice, and encrypted backups purge on a standard 30-day rotation. The full mechanics are in meeting data residency.
The vendor chain is public — on purpose
The honest version of the training question is really about subprocessors: the speech engine and language model your tool quietly forwards audio to. We wrote a whole guide on how to check any meeting tool's answer; this page is our own answer to the same questions.
So we publish the chain. Our subprocessors list names every third party that processes customer data on our behalf, with its purpose and processing location, and it mirrors the table in the DPA so the public page can never drift from the agreement.
The list isn't just informational — you have rights attached to it. We give at least 30 days' notice before adding or replacing a subprocessor, you can object on data-protection grounds, and if we can't resolve the objection you can terminate with a pro-rata refund. We also impose data-protection obligations equivalent to our DPA on each subprocessor and remain liable for their performance.
Verify it — don't take our word
Everything above is checkable in about ten minutes:
- Subprocessors — who touches your data, for what, and where.
- Data Processing Agreement — the retention table (including the no-training line on meeting audio), breach notification within 24 hours, deletion terms, and Standard Contractual Clauses for transfers.
- Security overview — how the encryption and access controls work.
- Legal hub — the current versions of everything, in one place.
If you're running a formal review, those four links are the packet. And if your policy says meeting content can't leave your infrastructure at all, Enterprise plans support self-hosting the entire stack — the strictest answer to the training question is that the data never reaches anyone's cloud, including ours.
Why we set it up this way
Sageio translates live meetings — sales calls, board discussions, HR conversations. People say things in meetings they would never paste into a chatbot, and letting an AI bot join a meeting is an act of trust we don't get to fumble.
Training on customer meetings would be a fast way to make the product marginally better and the trust permanently worse. We chose the boring architecture instead: audio in memory, text under your control, vendors under contract, and the paperwork public.
Frequently asked questions
Does Sageio use my meeting content to train AI models? No. Sageio does not train models on your meeting content, and its DPA records the commitment contractually: meeting audio is processed in memory only, never written to storage, and not used for model training. The text transcript and summary are retained only until you delete them.
Do Sageio's AI vendors train on my meeting audio? No. The speech-to-text and translation vendors that process your audio do so under documented no-training terms, and Sageio publishes its full subprocessor list so you can verify each vendor yourself. Sageio imposes data-protection obligations equivalent to its DPA on every subprocessor and remains liable for their performance.
Is my meeting audio stored anywhere? No. Audio is streamed for live transcription, processed in memory, and discarded — it is never written to storage. Only the text transcript, translations, and AI summary are kept, encrypted, in the region you choose (US, EU, or APAC), until you delete them.
Can Sageio identify people by their voice? No. Sageio does not use meeting content to identify any person by voice and creates no biometric templates — that language is in the Data Processing Agreement, not just this page.
What if my company can't send meeting data to any cloud? Enterprise customers can self-host the entire Sageio stack, so meeting content never leaves their own infrastructure. For everyone else, every plan starts with a free 60-minute trial (no credit card), so you can run one real meeting and review the documents before deciding anything.
The training question deserves a mechanism, not a slogan — and now you've seen ours. If you're evaluating tools more broadly, the practical guide to real-time translation for remote teams covers the rest of the decision. Or skip ahead: add Sageio to one real meeting on Google Meet or Microsoft Teams, watch the captions land in everyone's language, then delete the transcript and confirm it's gone. That's the whole point — it's yours.