Is Audio Transcription Private? Consent, Storage, and What Happens to Your Recordings

You are about to upload a recording of a meeting, an interview, or a therapy session to a website, and a reasonable question stops you: who will hear this? Where does it go? Who keeps a copy, and for how long? A recording is not like a document. It carries voices — and voices identify people, reveal emotions, and capture things said in confidence. Uploading one deserves a straight answer.

This guide gives it: what happens to a recording when you transcribe it online, what the law says about recording other people, and how to judge whether a transcription service is handling your audio responsibly. It also explains what AudioToText does with your files, in plain terms.

Why Audio Privacy Is Different

A document can be redacted. A recording cannot, not easily. A transcript can have names removed, but the audio still carries the voice, and a voice is a biometric identifier — it can identify a person as surely as a fingerprint. A recording also captures things a written record never would: the hesitation before an answer, the tone of a disagreement, the background conversation at the start of a call. People say things on audio they would never write down, and they often forget the recording is running.

That is why audio deserves more care than text. A leaked document is bad. A leaked recording of a private conversation is worse, because it carries the people themselves — their voices, their emotions, and the things they said when they thought nobody was keeping a record. Any service that handles recordings should be judged by how seriously it takes that difference.

What Happens to Your Recording When You Transcribe It Online

The path is the same for most services, and it is worth knowing in outline.

Upload. Your file travels from your device to the service's server, usually over an encrypted connection (HTTPS). If the connection is not encrypted, anyone on the network can intercept the file — a basic check is that the site address starts with https://, and any service that accepts uploads over plain HTTP should not be trusted with recordings.

Processing. The server decodes the audio and runs it through a speech recognition model. With most services, that model runs on the provider's own servers. With some, the audio is sent on to a third-party provider — a cloud speech service, for example. That distinction matters: if the audio goes to a third party, the privacy policy of that third party applies too, and you should be told about it.

Storage. The recording and the transcript are stored for some period — hours, days, or indefinitely, depending on the service. Some services delete files automatically after a set time. Others keep them until you delete them, or forever. The retention period should be stated plainly, not buried in a policy.

Deletion. When a file is deleted, it should be gone — from the live system and from backups within a reasonable time. A service that says "deleted" but keeps the file in a backup for a year is not being straight with you, and its policy should say how long backups keep deleted files.

Human review. Some services have people listen to recordings to improve their models or to check quality. If a service does this, it should say so clearly, and it should let you opt out. A recording of a private conversation should never be listened to by a stranger without your knowledge.

What AudioToText Does With Your Files

This is our policy, stated plainly:

• Your files are processed on our own server. We do not send your recordings to a third-party transcription provider. The audio stays on the machine that transcribes it.
• We never use your recordings to train models. Your audio is not training data, and it never becomes part of any dataset.
• Files are deleted automatically after 72 hours. Uploads and transcripts are removed 72 hours after processing, without you having to ask. You can delete any result immediately, from the result page or your account page, and the files are removed at once.
• We never publish, index, or share your files. Your recordings and transcripts are not added to any shared database, not indexed by search engines, and not shared with advertisers or data brokers.
• No human listens to your recordings. Transcription is automatic. Our staff only open a file if you ask us to look at a specific run — for example, if you report a problem through the feedback page — and then only that run.
• Completion emails carry only the transcript text, sent to the address you gave us, and only when you asked for it.

If any of that changes, we will say so on the privacy page before it takes effect. A privacy policy that changes silently is not a policy.

Why Consent Matters — and What the Law Says

Recording other people without their agreement is not only rude; in many places it is unlawful. The rules fall into two broad groups.

One-party consent. In many U.S. states, and in some other countries, it is lawful to record a conversation if one participant consents — which can be you. Even where this is the legal rule, telling people is still the right thing to do: a recording people do not know about destroys trust when they find out, and it can still be unusable as evidence or as a published record.

All-party consent. In other U.S. states (including California, Florida, and Illinois), and in many European countries, everyone being recorded must consent. Recording without that consent can be a criminal offence, and the recording may be inadmissible and unusable. The European General Data Protection Regulation (GDPR) adds another layer: a voice is personal data, and processing it requires a lawful basis — usually consent — plus proper handling, storage limits, and the right to deletion.

The practical rule is simple, and it covers both cases: always tell people you are recording, and get their agreement, before you start. Say it out loud at the beginning — "I'm recording this so we get accurate notes, is that okay?" — and do not record if anyone objects. For interviews and research, get consent in writing as well, and keep it with the recording. The U.S. Department of Justice publishes guidance on recording consent rules; see justice.gov for the federal view, and check the rules in your own state or country before you record.

How to Judge a Transcription Service's Privacy

Before you upload a sensitive recording to any service, check five things. They take ten minutes, and they tell you almost everything about how seriously the service takes privacy.

If a service fails two or more of these checks, do not upload sensitive material to it. Use it for public material only, or find a service that passes. The checks are also a good way to judge our own service — we publish the answers on our privacy page, and we invite the comparison.

How to Protect Sensitive Recordings

Record only what you need. If a meeting has a sensitive first ten minutes and a routine rest, record the routine part and take notes on the sensitive part. Less recorded means less to protect.

Trim before you upload. If a recording contains a passage that should never leave your device — a password spoken aloud, a private aside — cut it out before uploading. Most audio editors can do this in a minute, and it is far safer than trusting a service to delete it later.

Use the shortest retention you can. Delete the result as soon as you have the transcript and have used it. Do not leave recordings sitting in an account for months because you might need them one day. If you need the transcript later, keep the text, not the audio — text is easier to store securely and easier to redact.

Be careful with shared accounts and shared links. A transcript shared by link is a transcript anyone with the link can read. Share the text with named people, not with a public link, and check who has access when the work is done.

Mind the metadata. Audio files can carry metadata: the device that recorded them, the time, and sometimes the location. Strip metadata before uploading if it matters, and be aware that a transcript file carries the date and the file name you gave it.

Think about who is in the recording. A meeting transcript names the participants. An interview transcript names the interviewee. A medical or legal recording may contain information about people who never consented to be recorded at all — a patient mentioned by name, a client discussed in passing. Those people deserve the same care as the people in the room, and their details should be removed from any transcript you share. For more on handling interviews carefully, read our guide to transcribing interviews.

Common Mistakes to Avoid

Recording without telling anyone. The most common mistake, and the one with the worst consequences. Tell the room, get agreement, and do it on the record.

Uploading a recording you have not listened to. A recording can contain far more than you remember: the conversation before the meeting started, a phone ringing with a private call, a colleague speaking in the background. Listen to the first minute before you upload, and trim anything that should not leave your device.

Keeping recordings indefinitely. "I might need it one day" is how private recordings accumulate into a liability. Keep the transcript, delete the audio, and keep the transcript only as long as you need it.

Trusting a privacy policy you have not read. If a service will not say plainly how long it keeps files, whether it trains on them, and who can listen to them, it has answered your question by not answering it.

Sharing a transcript more widely than the recording. A transcript is easy to forward, copy, and paste — easier than audio, and harder to take back. Before you share one, check that everyone named in it would be comfortable with the audience you are sharing it with. See pricing for the plans — the Free plan covers 60 minutes a month, and every plan deletes files after 72 hours.

Frequently Asked Questions

Can the service hear my recording?

With AudioToText, no person listens to your recording. Transcription is automatic, on our own server, and our staff only open a file if you ask us to look at a specific run. Other services differ — check their policy for human review before you upload anything sensitive.

Is my recording used to train AI models?

Not by us, ever. Some services do use uploaded audio for training, sometimes with an opt-out and sometimes without one. The policy should say so plainly; if it does not, assume the worst and do not upload sensitive material.

How long do you keep my files?

72 hours, automatically. You can delete any result immediately from the result page or your account page. Backups do not include user files.

What if I record someone by accident?

Delete the recording. If you have already uploaded it, delete the result immediately and do not share the transcript. If the person asks what happened, tell them honestly.

Is transcription legal for medical or legal recordings?

It can be, but the rules are stricter: medical recordings may be protected health information, and legal recordings may be privileged. Get proper consent, use a service that meets the relevant standard, and check the rules that apply to your profession before you upload.

Key Takeaways

• A voice is personal data. Treat a recording with more care than a document.
• Always tell people you are recording and get their agreement, on the record, before you start.
• Judge a service by five checks: retention time, training use, third parties, human review, and deletion.
• Trim sensitive passages before uploading, delete the audio once the text is filed, and keep the transcript only as long as you need it.
• Share a transcript only with the audience everyone named in it would accept.

!Audio privacy checklist

Image description: The image should show a shield icon in the centre, with five small checkmarks around it labeled "Auto-delete in 72 hours", "No training on your audio", "No third-party processors", "No human listening", and "Delete any time". A caption reads "Five checks before you upload a sensitive recording."

References

• U.S. Department of Justice — recording consent and privacy guidance: https://www.justice.gov/
• U.S. Department of Health and Human Services — health information privacy: https://www.hhs.gov/
• Library of Congress — personal audio collections and preservation: https://www.loc.gov/