Privacy

We never receive your conversation. Reading the archive, hashing it, laying it out and writing the PDF all happen on your own machine, in the page you have open. Nothing about the chat is sent to us, because there is nowhere to send it: ChatExhibits is a set of static files, with no server that accepts uploads, no database and no log of what you processed.

A consequence worth stating plainly: we could not hand your conversation to anyone who asked for it, lawfully or otherwise, because we do not have it.

Five small things are kept in your browser’s local storage so the software is usable across visits. They never leave the machine, and you can clear them at any time through your browser’s site-data settings.

• chatexhibits.entitlements: proof that you paid for an export: the checkout reference, the tier, the digest of the archive it covers, and the date it stops being regenerable. This is what lets you regenerate free for 30 days without an account. For an Attorney subscription it holds the end of the period paid for, and when that date passes this browser sends the checkout reference, and nothing else, to our verification address to ask Stripe whether the subscription is still active.
• chatexhibits.pending: while you are away at Stripe, the SHA-256 digest of the export you are buying, so that when you come back we know which one your purchase covers. Stripe can only return a checkout reference to us, so this is written here instead of travelling with you. It is a hash of a file, not of a conversation, and it is deleted as soon as the purchase is recorded.
• chatexhibits.matters: the matters you create, each with its name, its Bates prefix, its counter width and where the next production starts, so numbering stays continuous.
• chatexhibits.firm: the letterhead you typed, to print on the certificate.
• chatexhibits.bates: the numbering prefix, its width and the number your next production should start at, so a second document resumes instead of reusing numbers the first one already used.

No message, media file or transcript is ever written to any of them. A test in the source refuses to build if that changes, and the same test bans IndexedDB and session storage outright rather than trusting the convention.

The site sets no cookies and carries no advertising or social tracker. To know how many people visit and which pages they read, it counts page views with PostHog, hosted in the European Union. What is sent is the page address without its query string, the referring site, your browser and your device type. PostHog counts unique visitors from a hash that changes every day and discards your IP address before anything else reads it, so nothing locates you, nothing identifies you across days, and nothing is stored in your browser. It also counts a few steps of the tool, the same way: the dialog opened, an export read or refused, a preview built, a checkout button pressed, a payment confirmed. Each carries at most the layout, the tier or the billing period, and never a file name, a participant, a message or a figure from your export.

Nothing about your conversation is ever part of it. Session recording, click tracking and heatmaps are switched off in the source, so what appears on screen while you preview an export is never captured. This is why you were never shown a consent banner: nothing is stored on your device and nothing profiles you.

Loading a page means requesting files from Cloudflare, Inc., which serves them. Like every web server it keeps ordinary request records (IP address, time, the file requested, the browser identifying itself) for a limited period, for security and to keep the service running, under its own policies as our processor. That is true of every website you visit, and it never includes anything about the contents of your chat.

Payment is handled entirely by Stripe on Stripe’s own pages. We never see, receive or store your card number: what comes back to us is a checkout reference and whether it was paid, and for a subscription whether it is still active and until when. Stripe processes your payment and contact details as its own controller, under its own privacy policy, and it is the only third party involved in a normal purchase.

Voice notes are transcribed on your own machine. A speech model runs in your browser, in a background thread, and the audio is never sent anywhere. There is no option to send it, no checkbox to tick and no accuracy tier that changes this: the promise covers the audio exactly as it covers the text.

The first transcription downloads the model itself, around eighty megabytes, from the public repository that publishes it. That request reveals your IP address to that host, the way requesting any file from any server does, and it carries nothing about your conversation: it asks for a public file by name. The model is then cached in your browser and never fetched again.

A chat export is mostly other people’s words, and often their photographs, voices and phone numbers. We never receive any of it, so we are not processing it on your behalf. Whether you may use that material, in a filing or anywhere else, is between you and your lawyer, and remains your responsibility.

Where the law gives you a right of access, correction, erasure or portability, it applies to personal data we hold. For your conversation, there is none. What exists is your purchase record at Stripe and, if you write to us, the email itself. Write to martin@chatexhibits.com and we will answer.

Data controller: Martin Ferret EI, 9 rue des Érables, 45250 Briare, France, SIREN 999 111 438. You may also complain to the CNIL, the French data protection authority.

You do not have to take our word for it. Open your browser’s network panel and use the site: no request carries your archive. The digest checker and the export viewer work with your network disconnected, which is the shortest proof there is.