November 1, 2026 · 13 min read
Captioning Privacy: What Happens to Speech and Transcripts
What live captioning tools collect, store and share, which U.S. consent rules apply, and a checklist for churches and event teams to protect caption data.
Live captions turn everything a speaker says into text, so the privacy question is fair: where does that speech go, who can see the text, and does anyone keep it? The answer depends on the tool. Some captioning systems keep audio, transcripts and attendee details. Others keep very little. Before you caption a service, class or event, find out what your tool collects, how long it keeps it, and whether the vendor can use it to train AI. Then tell your audience in plain words.
This guide covers what captioning systems can collect, the U.S. consent rules that matter, a checklist for reading a vendor’s policy, and how Live Caption AI handles the data it touches.
TL;DR
- Captioning tools can collect more than words: audio, transcripts, names, participation data and timing data. What they keep varies by vendor.
- Some AI vendors retain data, share it with third parties, or use it for model training. Read the license terms and restrict training use in writing.
- U.S. recording consent differs by state. Federal law generally needs one party’s consent, and some states require everyone’s.
- Tell attendees what is captured, why, who can access it, and how long it is kept.
- Live Caption AI never stores session audio. It is an AI captioning tool, not a certified or human-reviewed service, and it is not HIPAA-compliant.
Table of Contents
- What a captioning system can collect
- Are captions and transcripts stored, and who can see them?
- Can vendors train AI on your speech?
- U.S. consent and privacy rules in plain terms
- How to read a captioning tool’s privacy policy
- How churches and event organizers can protect caption data
- How Live Caption AI handles your data
- Our perspective
- Translation and privacy
- Explore Live Caption AI
- A simple privacy notice you can adapt
- FAQ
- Sources
What a captioning system can collect
Captions look harmless. A line of text on a phone screen. Behind it, though, a system may handle several kinds of data:
- Audio. The speech itself, sent to a speech-to-text engine.
- Transcripts. The text, which can be saved, exported or summarized.
- Attendee data. When captions run inside a meeting platform, providers may process names, usernames, profile photos, email addresses and participation data such as audio, video and chat. One health insurer’s meeting-recording notice lists exactly these categories.
- Context data. One major platform’s captioning may use a meeting’s subject, invitation, participant names and attachments to improve accuracy.
- Performance data. Some captioning systems gather timing and cadence data to synchronize future events, according to one patent filing.
None of that is automatically a problem. It becomes one when nobody told the people in the room.
Audio versus text
The two carry different risks. Audio holds a voice, which is personal. Text holds names, prayer requests, health details or financial figures, which can be sensitive. A tool that discards audio right after converting it removes one of the two. The text still needs a plan: who can export it, where the file goes, and when you delete it.
Are captions and transcripts stored, and who can see them?
It depends entirely on the product. One university’s help page says that a large meeting platform permanently deletes caption data after a meeting ends. Other tools keep recordings and transcripts in an account until someone removes them.
The extreme case shows why defaults matter. In a 2023 complaint, the FTC and DOJ alleged that a voice assistant’s default settings kept voice recordings and transcripts indefinitely and that some deletion requests were not honored. That was a smart speaker, not a captioning app. The lesson carries over: the default setting is what most people end up living with.
Ask three plain questions of any tool:
- Is the audio kept after the words are transcribed?
- Is the transcript kept, and for how long?
- Who in our organization, and at the vendor, can open it?
Can vendors train AI on your speech?
Sometimes, yes. A law firm’s April 2026 analysis of AI transcription tools notes that vendors may retain data indefinitely, share it with third parties or use it for model training, depending on their license terms. The same analysis advises organizations to restrict training use contractually unless people have consented.
For a church, this matters more than it first seems. A sermon is public. A prayer-request time, a counseling-adjacent talk or a members’ meeting is not. If speech from those moments feeds a vendor’s model, your members were never asked.
Look for this in the terms of service and privacy policy, not in marketing copy. Search for the words “train,” “improve our services,” “third party” and “retain.”
U.S. consent and privacy rules in plain terms
This is general information, not legal advice. If your situation is complicated, ask an attorney licensed in your state.
Recording consent
Captioning is not always recording, but when audio or transcripts are saved, wiretap and recording laws can come into play. According to an employment-law firm’s overview, federal law generally allows recording when at least one person knows and consents, provided the purpose is not criminal. Some states require all parties to consent. The same overview points to California privacy laws and Illinois’ biometric privacy law (BIPA) as other rules that may apply to AI recording and transcription.
If you host events in several states, or have remote attendees, follow the strictest rule that could apply. Announcing that captions and transcripts are running costs nothing. If you also need written permission for photos or recordings of minors, a set of consent form templates can give you a starting point for the wording.
Consumer privacy laws
State privacy laws such as the CCPA call for transparent data practices and give people rights to delete, access and opt out, as that same law-firm analysis explains. Whether a law covers your organization depends on its size, its type and the data involved. Many churches and small nonprofits fall outside some of these laws. Your vendor may not.
Health information
If you caption a medical setting, HIPAA enters the picture. HHS guidance says that when a covered entity uses a cloud service provider to handle electronic protected health information on its behalf, the provider is a business associate and a business associate agreement is required. Live Caption AI does not offer a BAA and is not HIPAA-compliant. If protected health information may be spoken, talk to your compliance officer before choosing any captioning tool.
How to read a captioning tool’s privacy policy
You do not need a lawyer to run a first pass. Use this checklist, based on what the law-firm analysis recommends assessing: data handling, retention and deletion, training and third-party sharing, encryption and access controls, and the ability to delete data on request.
| What to check | What a good answer looks like |
|---|---|
| Audio retention | Stated plainly: kept or not kept, and for how long |
| Transcript retention | You control it, or a clear deletion schedule exists |
| Model training | No training on your data without your consent, in writing |
| Third-party sharing | Named subprocessors, or a clear statement of none |
| Access controls | Only authorized people can open or export transcripts |
| Deletion on request | A real process, with a contact |
| Attendee data | No account or sign-up needed to read captions |
One more point: retention policies should line up with your legal-hold duties and delete obsolete data regularly. If you keep AI transcripts, treat them like any other record, as the same analysis advises.
How churches and event organizers can protect caption data
The employment-law overview recommends several steps that translate directly to a congregation or conference. Its guidance, for recording and transcription notices, is to give advance written notice, use a verbal announcement where possible, limit access to authorized people, and bind vendors with written data-protection agreements.
Here is what that looks like in practice.
Before the event
- Write a short notice. Say what is captured (live captions, and a transcript if you save one), why (accessibility), who can access it, and how long you keep it.
- Put it where people look. Next to the QR code, in the bulletin, on the registration page.
- Pick the lowest-data setup. If you only need live captions, do not save a transcript.
During the event
- Announce it. One sentence from the speaker works: “Live captions are available. Scan the code to read along.”
- Keep sensitive moments out. Pause the session for prayer requests, personal testimonies or closed-door discussion if you do not want them transcribed.
After the event
- Decide what you keep. A sermon transcript you want to publish is different from a board meeting you do not.
- Limit who can export. Give transcript access to the people who need it.
- Delete on schedule. Set a retention date and stick to it.
How Live Caption AI handles your data
Here is what we can say about our own product, and nothing more than that.
- Session audio is never stored.
- Attendees need no app and no account. They scan a QR code and read captions on their own phones. No sign-up means no attendee profile to protect.
- You control the transcript. Transcripts can be saved and exported as TXT, SRT, JSON or PDF. If you do not save one, you have nothing to manage.
- You choose the setup. The speaker runs it from a phone or laptop. No hardware to rent. No stenographer to book.
What Live Caption AI is not: it is an AI captioning tool, not a certified or human-reviewed service. It does not make an organization ADA-compliant by itself, though an organization can use it as part of its ADA effective-communication plan. It is not HIPAA-compliant, it does not offer a BAA, and it should not be treated as suitable or certified for courtroom or clinical use.
Accuracy is a privacy issue too
A caption that garbles a name or a diagnosis is a small data-quality failure, and it can mislead the reader. That is why we built specialized models: Medical, Worship & Ministry, and Finance & Earnings. You can also add up to 100 Custom Terms, the names, places and terms you type in before going live. Better accuracy means fewer corrections afterward, and fewer people editing saved transcripts.
Our perspective
Most captioning privacy problems start with a default nobody chose. A tool keeps audio because that was easy to build. A transcript sits in an account because nobody set a deletion date. An attendee is asked to create an account because the vendor wanted a mailing list.
We built Live Caption AI to avoid that pattern. Session audio is never stored. Attendees scan a QR code and read captions on their own phones, with no app and no account. You decide whether a transcript exists at all, and you export it only when you want to.
We are also direct about limits. We are an AI captioning tool, not a certified or human-reviewed service. We are not HIPAA-compliant and we do not sign BAAs. If you are a church, a school, an AV team or an event organizer who wants live captions on every phone in the room at a price a small budget can carry, that is the job we do well. Tell your audience what is captured, keep what you need, delete the rest.
Ryan
Translation and privacy
Live translation adds another step in the data path, so say so in your notice. Live Caption AI translates into one of 29 languages per session: the speaker picks one source language and one target language, and viewers can switch between the original and the translation. For a multilingual service or conference, that is one more reason to post the notice in the languages your audience reads. If you save the transcript, the export is text only.
Explore Live Caption AI
If you want captions on every phone in the room without hardware, accounts or stored session audio, look at the plans. Professional is $19.99 a month with unlimited captioning. Business starts from $199 a month. A Free plan is available too, and one-off translation credit packs run from $6.99 for 30 credits to $19.99 for 120. For medical, legal and professional settings, read how those teams use it on the professional page.
No hardware. No stenographer. No app for attendees to install.
For more guides on live transcription, browse our real-time transcription posts.
A simple privacy notice you can adapt
Short notices get read. Adapt this wording to your setting and review it with your counsel if you operate in a state with all-party consent:
Live captions are available today. Scan the QR code to read them on your phone. Captions are generated by AI and may contain errors. [We do not save a transcript / We save a transcript for [purpose] and keep it for [period]. Contact [name] to ask questions or request deletion.]
Keep the bracketed parts honest. If you save transcripts, say so. If someone asks for deletion, have a person who can act on it.
FAQ
Does live captioning record my voice?
Not necessarily. Live captioning converts speech to text as it happens. Whether audio is kept depends on the tool. Live Caption AI never stores session audio. Always check the policy of any tool you use.
Do I need permission to caption a service or event?
Captions alone are usually less sensitive than recordings, but laws differ by state, and saving transcripts or audio can trigger recording and privacy rules. Federal law generally needs one party’s consent, and some states need everyone’s. Announce captioning in advance and ask an attorney if you are unsure.
Can captioning vendors use my speech to train AI?
Some can, depending on their terms. Review the license and privacy policy for words such as train, improve, retain and third party. If training use is allowed, ask for it to be restricted in writing. For Live Caption AI, we state that session audio is never stored.
No. Live Caption AI is not HIPAA-compliant and does not offer a Business Associate Agreement. It is not suitable or certified for clinical use. If protected health information may be spoken, speak to your compliance officer before choosing a captioning tool.
Sources
- Privacy Notice: Recording and Transcription of Meetings (Centene)
- Teams Meetings Real-Time Captions FAQ (University System of New Hampshire)
- US10692497B1: Synchronized captioning system and methods
- FTC and DOJ complaint against Amazon (2023)
- AI Transcription Tools Under Scrutiny: Navigating Privacy (Goodwin, April 2026)
- The legality of AI-powered recording and transcription (Reed Smith)
- Guidance on HIPAA & Cloud Computing (HHS)