How we handle your data

No legal jargon. Here is exactly what happens to your files when you upload them to DataSharpener — and what never happens.

Last updated: April 15, 2026

When you upload a file

Your file is processed in memory — never written to disk

When you upload a CSV or Excel file, it is parsed entirely in your browser (client-side). The raw file bytes never leave your machine and are never stored on our servers.

We store column statistics — not your data

We save metadata about your dataset: column names, types, min/max/mean/median for numeric columns, top 8 most frequent values for categorical columns, null counts, and health scores. This is what powers your charts and insights.

We store a 500-row preview — for charts and AI Q&A

To render your data preview table and to let the AI answer questions, we save the first 500 rows of your dataset. If your file has fewer than 500 rows, we store all of them. This preview is stored securely and tied to your account.

Q&A conversations are saved per dataset

When you use "Ask the Data", your questions and the AI answers are saved so you can scroll back through your session history. You can clear this history at any time with the "Clear history" button.

What we never do

We never store your full dataset

Your complete file — every row, every column — is never persisted anywhere. Only the statistics and 500-row preview described above are saved.

We never sell or share your data

Your uploaded files, statistics, and conversation history are private to your account and are never sold, rented, or shared with any third party for any purpose.

We never use your data to train AI models

When you ask a question, a small context window (your column names, stats, and up to 500 preview rows) is sent to Google Gemini to generate an answer. Google's API terms prohibit using API inputs for model training.

We never access your data without your action

No DataSharpener employee can browse your datasets. Access requires your authenticated session. There are no admin backdoors to user data.

Data lifecycle

Upload

File parsed in-browser → statistics + 500-row preview sent to our API → stored in your private account.

Analysis

Charts, insights, and Q&A use only the stored statistics and preview. Your original file is gone.

Retention

Data is kept until you delete it. There is no automatic expiry on the free plan.

Deletion

Click the trash icon on any dataset card. This permanently deletes the statistics, preview rows, and Q&A history from our database. There is no recovery.

Access & isolation

Every request is authenticated

We use Clerk for authentication. Every API call that reads or writes your data requires a valid session token. Unauthenticated requests are rejected before they reach the database.

Row-level security isolates every user

Our database (Supabase) enforces row-level security on every table. Even if two users share the same database, each user's queries are filtered to only their own rows. It is not possible to read another user's data.

All data in transit is encrypted

Every connection to DataSharpener is over HTTPS (TLS 1.2+). Data at rest in Supabase is encrypted by AES-256.

Third-party services

We use a small number of sub-processors. Here is exactly what each one receives.

Supabase

Database

Stores your column statistics, 500-row preview, and Q&A history. Hosted on AWS (us-east-1 by default). SOC 2 Type II certified.

Clerk

Authentication

Handles sign-up, sign-in, and session management. Receives your email address and name. Does not receive any of your uploaded data.

Google Gemini

AI (Q&A + summary)

Receives your column names, column statistics, and up to 500 preview rows when you use AI features. Google's API terms prohibit using this data for training. No data is stored by Google beyond the request.

Vercel

Hosting

Serves the application. Receives standard HTTP request metadata (IP, user-agent, path). Does not receive your uploaded data or database contents.

Questions about your data?

Email us at privacy@datasharpener.com and we will respond within 2 business days.