I'm nosy, not careless.

In one breath

You give me a chat export. I read it once, write your verdict, and delete the raw conversation. I keep the verdict you paid for — never the messages. I don't train any model on your chats, and I don't sell or share them. The file travels encrypted, and no human at Margot reads it.

What I collect

• The chat export you upload. Used only to generate your verdict, then deleted from our servers. It is processed in memory for the length of one request and not written to a database.
• Your generated verdict. The rendered report is held in your own browser session so the page can display it. We do not store a copy tied to your identity.
• Payment details. Handled entirely by Stripe. We never see or store your card number — only a payment reference confirming the charge succeeded.
• Basic technical data. Standard request logs (IP, timestamp, error messages) for security and abuse-prevention. These never contain the contents of your chat.

What I never do

• Train any AI model on your conversations — not ours, not anyone else's.
• Sell, rent, or share your chat or verdict with third parties for marketing.
• Keep your raw messages after the verdict is generated.
• Read your chat by hand. The analysis is automated.

Who processes your data

Two sub-processors make Margot work: Anthropic (the AI that writes the verdict — sent your chat text for the single generation, governed by their commercial terms, which do not train on API inputs) and Stripe (payments). That's it.

Your choices

Because we don't keep your chat or tie verdicts to an account, there's little standing data to request or delete. If you emailed us or paid, you can ask us to remove those records: hello@margot.app. Everyone in a group chat should be comfortable with it being read before you upload it — that's the consent box on the upload screen.

Changes

This policy may be updated as the product grows. Material changes will be reflected here with a new date. Questions: hello@margot.app.

Last updated: 24 June 2026.