Data Usage Policy

1. On-device processing

When you load a dataset or connect a database, the ATARS desktop app processes it locally on your own computer to run the analysis you request — profiling, cleaning, charting, modelling, forecasting and report generation. Nothing is uploaded to us; we operate no data servers.

2. We don't sell or share your data

Your datasets and their contents are never sold, rented or shared with advertisers or unrelated third parties. They exist to power your analysis, nothing else.

3. Full datasets, no silent limits

ATARS is designed to load your data fully and honestly. It does not silently cap or truncate rows behind your back; where practical limits exist (your computer's memory), they are surfaced to you rather than hidden. For very large databases, SQL Lab can push aggregation to the database itself.

4. What can leave your computer, and when

ATARS runs completely offline by default. Something leaves your computer only when you switch on a feature that needs the internet, and then only what that feature needs:

• A cloud AI provider — the portion of a request needed to answer it goes to that provider only. For text columns, themes are computed locally and only a few representative words are sent to be named, never your rows. You choose the provider and remain bound by its data terms.
• A local AI model — nothing leaves; the model runs on your machine.
• Web search — the text of the question you allow to use the web goes to DuckDuckGo. Your dataset does not.
• The update check — off by default; when used, it fetches a small signed version file and sends nothing from your analysis.
• Matching documents by meaning — off until you switch it on. The first time, ATARS downloads a small language model (about 90 MB) to your computer; after that it runs entirely offline. Your documents and your questions are never sent anywhere — only the model comes down.

5. Documents, text and your .atars files

Documents you add (policies, contracts, past reports) and the search index built from them are stored on your own computer, in your ATARS folder, so they are remembered between sessions. An .atars file contains your data together with its cleaning, charts, notes, metric definitions and receipts. When you send one to someone, they receive the data inside it — share it as carefully as you would share the dataset.

6. Database & warehouse connections

When you connect to a database or warehouse — SQLite, MySQL, PostgreSQL, Amazon Redshift, Snowflake or Google BigQuery — ATARS reads the data you query to perform analysis. Credentials are used only to establish the connection you request and are not stored in the application's source code.

The terminal edition has no database connections at all — it reads files only: CSV, Excel, JSON, Parquet, Feather and TSV.

7. Exports

Reports and files you export (PDF, DOCX, PPTX, Excel, Markdown, a runnable Jupyter notebook, CSV, JSON and images) are generated for you to download and keep. Reports include a record of where their numbers came from, re-checked against your data when the report is made. Once downloaded, they are under your control and responsibility.

8. Your responsibilities

• Ensure you have the right to upload and analyse the data.
• Avoid uploading sensitive personal data unless you have a lawful basis to process it.
• Handle exported files securely.
• Comply with the data-protection laws that apply to you.

9. Questions

For anything about how your data is handled, contact us through the contact page.