Question · Voice privacy and security

Does speech to text share your data with third parties?

Sometimes. Dictation sends your audio to a third party only when the recognition runs on a server or the app passes it to another company, such as a cloud provider or an AI model maker. Recognition that runs on your device sends the audio nowhere, and the app's own policy still decides what else it shares.

Updated October 17, 20265 min read8 cited facts, each traced to its source

Sometimes. Whether dictation shares your words with anyone else depends on one question: where the recognition runs. If on-device recognition runs on your computer and the app sends nothing, no outside company receives the audio. If the audio goes to a cloud dictation server, the company behind the app receives it, and that company may pass it to cloud providers, labeling vendors or an AI model maker. Each of those hops is a third party with its own terms.

Where the recognition runs decides whether audio leaves your device

The table sorts the common setups by where the audio goes. The last column is the question to ask for each.

Setup	Where the audio goes	Who else can receive it
On-device recognition, such as Mac Dictation's general text mode ¹	Stays on the computer	No one for the audio itself. The app may still send account or usage data, so read its policy
An app using Apple's Speech Recognition permission ²	Possibly to Apple	Apple and its service providers ³
An app that calls a model company's API, such as Whisper To Input with an OpenAI key ⁴	To OpenAI's servers, or to a Whisper server you host	OpenAI, under its API terms, or no outside company if you host the server yourself
A cloud meeting or dictation service ⁵	To the vendor's servers	The vendor's listed service providers

The sharing question is therefore really a chain question. Ask who processes the audio first, then ask who that company hands it to.

Apple's answer depends on which speech interface an app uses

On a Mac, Apple says that voice inputs and transcripts for general text Dictation are processed on your device and are not sent to Siri servers ¹. That covers the built-in path, not every app.

Third-party apps are different. Apple's Siri, Dictation and Privacy page says that if you allow apps to use Speech Recognition for transcription, "the audio data to be transcribed may be sent to Apple" ². The same page says Apple "may process and store this information with trusted third-party service providers" ³. The passage quoted here does not name those providers.

The developer documentation adds a split inside Apple's own tools. The permission process applies to speech recognition with SFSpeechRecognizer, and the transcriber modules in SpeechAnalyzer do not send the user's voice audio to Apple's servers ⁶. So two apps that both use Apple's speech tools can behave differently. The app's developer chooses which interface to use.

AI dictation apps add a second company, and the model maker's terms apply

An app that transcribes with an AI model usually sends the audio to the model maker. Whisper To Input says its recognition runs on either the OpenAI API with your own key or a Whisper server you host yourself ⁴. In the first case, OpenAI is a third party to you and to the keyboard's developer.

OpenAI's own dictation works the same way. Its FAQ says that when you press the microphone icon in ChatGPT, the recorded audio is sent to its models to be transcribed ⁷. For developers, OpenAI says data sent to its API is not used to train or improve its models unless the customer explicitly opts in ⁸. That statement is about training. It does not tell you what the app that called the API keeps, and that is covered by the app's own policy. For the training side, see whether dictation apps train AI on your voice.

Cloud services list their vendors, and some vendors work on the data

Read the sharing section of a cloud service's policy, not only the training section. Otter's privacy policy says it discloses personal information to vendors and service providers, and one category is "Data labeling service providers who provide annotation services and use the data we share to create training and evaluation data" ⁵. That sentence means a third party can work on the data you send, not just store it. Whether your own recordings reach that vendor is a separate question that the policy answers only in general terms. For Otter specifically, see whether Otter.ai uses your recordings to train its AI.

It depends on five things, and each one can change the answer

Recognition location. On-device recognition keeps the audio on the computer. Server recognition sends it to whoever runs the server.
The speech interface. The same operating system can offer a path that sends audio and a path that does not, as the Apple split shows above.
The app's model provider. An app that calls an AI API adds that company to the chain.
The plan. Business and API accounts can have different terms from consumer accounts, so check the one you actually use.
Who the policy names. A policy that names categories of vendors is different from one that names none. A category such as "service providers" can still include a labeling company.

What to do

Find out where recognition happens. Check the app's settings for an offline or on-device option. On a Mac, check the Dictation settings under Keyboard, where Apple says you can confirm the on-device processing ¹.
On a Mac, review which apps may use Apple's Speech Recognition, and turn off the ones you do not need ².
For an app that calls an AI API, read the app's policy for what it keeps, then check the model maker's API terms for training and retention.
In any cloud policy, search for "service providers", "vendors" and "labeling". Those words tell you who else gets the data.
If you want audio to stay on your computer, pick an app that states it runs on the device. We make Yap, an on-device dictation app for Mac and Windows, so its recognition runs on your computer. Check its policy the same way you would check any other app.

The guides on whether voice to text is safe and voice data retention cover the steps after you know where the audio goes.

Evidence

Every number on this page, with where it comes from. Open a fact to see the exact words in its source.

Go deeper

Question · Voice privacy and security

Does speech to text share your data with third parties?

Updated October 17, 20265 min read8 cited facts, each traced to its source

Where the recognition runs decides whether audio leaves your device

The table sorts the common setups by where the audio goes. The last column is the question to ask for each.

Setup	Where the audio goes	Who else can receive it
On-device recognition, such as Mac Dictation's general text mode ¹	Stays on the computer	No one for the audio itself. The app may still send account or usage data, so read its policy
An app using Apple's Speech Recognition permission ²	Possibly to Apple	Apple and its service providers ³
An app that calls a model company's API, such as Whisper To Input with an OpenAI key ⁴	To OpenAI's servers, or to a Whisper server you host	OpenAI, under its API terms, or no outside company if you host the server yourself
A cloud meeting or dictation service ⁵	To the vendor's servers	The vendor's listed service providers

The sharing question is therefore really a chain question. Ask who processes the audio first, then ask who that company hands it to.

Apple's answer depends on which speech interface an app uses

On a Mac, Apple says that voice inputs and transcripts for general text Dictation are processed on your device and are not sent to Siri servers ¹. That covers the built-in path, not every app.

AI dictation apps add a second company, and the model maker's terms apply

Cloud services list their vendors, and some vendors work on the data

It depends on five things, and each one can change the answer

Recognition location. On-device recognition keeps the audio on the computer. Server recognition sends it to whoever runs the server.
The speech interface. The same operating system can offer a path that sends audio and a path that does not, as the Apple split shows above.
The app's model provider. An app that calls an AI API adds that company to the chain.
The plan. Business and API accounts can have different terms from consumer accounts, so check the one you actually use.
Who the policy names. A policy that names categories of vendors is different from one that names none. A category such as "service providers" can still include a labeling company.

What to do

Find out where recognition happens. Check the app's settings for an offline or on-device option. On a Mac, check the Dictation settings under Keyboard, where Apple says you can confirm the on-device processing ¹.
On a Mac, review which apps may use Apple's Speech Recognition, and turn off the ones you do not need ².
For an app that calls an AI API, read the app's policy for what it keeps, then check the model maker's API terms for training and retention.
In any cloud policy, search for "service providers", "vendors" and "labeling". Those words tell you who else gets the data.
If you want audio to stay on your computer, pick an app that states it runs on the device. We make Yap, an on-device dictation app for Mac and Windows, so its recognition runs on your computer. Check its policy the same way you would check any other app.

The guides on whether voice to text is safe and voice data retention cover the steps after you know where the audio goes.

Evidence

Every number on this page, with where it comes from. Open a fact to see the exact words in its source.

Does speech to text share your data with third parties?

Where the recognition runs decides whether audio leaves your device

Apple's answer depends on which speech interface an app uses

AI dictation apps add a second company, and the model maker's terms apply

Cloud services list their vendors, and some vendors work on the data

It depends on five things, and each one can change the answer

What to do

Related questions

Evidence

Go deeper

Does speech to text share your data with third parties?

Where the recognition runs decides whether audio leaves your device

Apple's answer depends on which speech interface an app uses

AI dictation apps add a second company, and the model maker's terms apply

Cloud services list their vendors, and some vendors work on the data

It depends on five things, and each one can change the answer

What to do

Related questions

Evidence

Go deeper