Privacy and Data Ownership: Exploring Whisper Model's Data Flow and OpenAI's Involvement

[mjund]

Hey everyone,

I have a question regarding the deployment of my own Whisper model. I wanted to clarify whether the data, specifically the voice recordings from my users or the text generated by the Whisper model, has any chance of being sent to OpenAI's servers.

I value privacy and data ownership, so it's crucial for me to understand the data flow and ensure that user data remains within the confines of my deployment. If anyone has insights or information on this matter, I would greatly appreciate your input.

Thank you!

[jwnacnud]

If you have concerns, try loading it onto your local machine and then disconnect from the internet. If it still works, then you can be sure it isn’t “phoning home”.

On Sat, Jun 3, 2023 at 11:24 AM mazahd ***@***.***> wrote: I have a question regarding the deployment of my own Whisper model. I wanted to clarify whether the data, specifically the voice recordings from my users or the text generated by the Whisper model, has any chance of being sent to OpenAI's servers. I value privacy and data ownership, so it's crucial for me to understand the data flow and ensure that user data remains within the confines of my deployment. If anyone has insights or information on this matter, I would greatly appreciate your input. Thank you! — Reply to this email directly, view it on GitHub <#1424>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGW5A7VIW57LBA7WB3Z7NLXJNXMNANCNFSM6AAAAAAYZM74HQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Jeffrey Duncan

[mjund]

maybe there is a background process like a whisper daemon or similar that handles the task of sending data if there is an available network connection, and if not, tries again later?

[phineas-pta]

the code is open source, you can read it

the model weights are available, you can inspect it

you can also monitor your connections to see which program connect owhich website

[jwnacnud]

If you are hyper concerned about making sure that your data never goes online, run this on a machine that is not connected to the Internet. You could also run it on a virtual machine that has no network connections.

On Sat, Jun 3, 2023 at 11:14 PM mazahd ***@***.***> wrote: maybe there is a background process like a whisper daemon or similar that handles the task of sending data if there is an available network connection, and if not, tries again later? — Reply to this email directly, view it on GitHub <#1424 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGW5A3F7XDAZ2MRP65DVETXJQKR5ANCNFSM6AAAAAAYZM74HQ> . You are receiving this because you commented.Message ID: ***@***.***>

-- Jeffrey Duncan

[eddywashere]

This question is referenced in other places as being an answer or at least a response to the question of data privacy. For folks who cannot review this codebase or modify the usage setup of whisper, would someone be kind enough to verify that there is no process for pushing data out to OpenAI?

example by @mjund

maybe there is a background process like a whisper daemon or similar that handles the task of sending data if there is an available network connection, and if not, tries again later

[ryanheise]

Also see: #1462 (reply in thread)