Just starting - question about privacy of data/code

Support Help
1

Hello everyone.

I am a retired senior who is just over 60 and found this tool and wanted to learn this.

I use to work for public sector, so security and privacy of anything we do is of the upmost importance/critical.

This is my understanding of how the Autocomplete works.

  • When a developer types, the Cursor client (the editor on the developer machine) collects a small snippet of the current code context, then encrypts this snippet locally before sending it over the network to Cursor’s cloud servers.

  • On the server, the snippet is securely decrypted, and Cursor’s in-house code LLM quickly generates a completion suggestion.

  • The predicted code is then returned to the client and displayed inline.

So my question is:

If I type any code, are any of my code stored on their servers? If so, where are the servers located?

I would guess that the LLM would need data for it to improved its auto suggest, correct?

I live in Canada (BC) so the privacy rules by the government here are very strict for public sector organizations (post secondary schools, gov’t branches, etc). They require that datacenters have to be in Canada.

Can anyone correct or help me understand this? I think I did not quite get it and open to correcting me.

Thanks

Dave

2

Hi @corestaples and welcome to Cursor Forum.

Your understanding of Autocomplete is correct.

With Privacy Mode setting in Cursor you can enable “Privacy Mode”.

When using Autocomplete that means no code from Autocomplete is permanently stored on the servers but only used for inference.

If someone wishes to contribute to Autocomplete improvements they may disable Privacy Mode. Only then is the data usable for training, but not when enabled.

Most data centers Cursor uses are in USA. I am not aware if AI providers have actual data centers in Canada but can inquire.

Here are the actual docs:

3

thank you

1 Like