How is the collection of private keys and secrets avoided when data sharing is enabled?

Situation

Private keys, secrets, passwords, etc. can often be found locally in files like .env files, appsettings.json and secrets.json, TODO.md or simply in uncommitted files.

Question

What measures have been taken to ensure those secrets are not included in the code snippets scraped for AI training?

Answer

For collected data, in terms of privacy, JetBrains filters and minimizes analytics data (for example, by applying regex and other filters) to reduce risk and avoid processing personal data.

We use special methods (functions and internal AI models) to clean secrets, passwords, etc. and replace them with placeholders.

Directories and files starting with a dot (.) are NOT included in data collection.

When data collection is enabled (starting with release 2025.3.1), we will apply deterministic and statistical filters to ensure that only data useful for improving model performance is collected.

We are also actively exploring options to let users turn off specific file types and directories, and we’ll share updates in the official blog as that evolves.