Privacy

1. Local processing

• Your .txt or .zip file is read in your browser. It is never uploaded to a server.
• From a .zip we only read the chat’s .txt; photos and videos are never opened.
• Every stat, leaderboard, award, emoji, word and phrase is calculated by code running on your device.
• The full report, the Wrapped, sharing and the PDF are free and are generated without sending anything anywhere.
• Once the analysis is done, the original text is thrown away. Only the aggregated stats and the pseudonymized version for the AI remain, in memory, until you close the page.
• We never store your chat’s content in a database or in your browser’s storage.
• The "Who said it?" game is only created if you ask for it: you choose which messages appear, and they are encrypted in your browser before a copy is stored on our server for 90 days. The key lives only in the link (after the #), which never reaches the server, so we cannot read it. Anyone with the link can read those messages and who wrote them, so you must confirm you have the group’s permission.
• Sharing the Wrapped with a link works the same way: the stats, the names and, if you choose, the AI analysis (with real quotes) are encrypted in your browser and a copy is stored on our server for 90 days. The key lives only in the link (after the #), which never reaches the server, so we cannot read it; if you don’t generate a link, nothing is stored. Anyone with the link can see it, so you must confirm you have the group’s permission.

2. Phone numbers

• Participants who show up as a phone number are displayed as “Contact 1”, “Contact 2”…
• Phone numbers written inside messages are removed before any analysis.

3. What is sent to the AI (only if you turn it on)

The AI analysis is optional: a profile of each person (role, superpower, kryptonite, catchphrases and best quote), the group’s big moments with dates, made-up awards backed by the quote that proves them, broken promises, eternal debates, the group dictionary, plus topics, your story, emotions and relationships, inside jokes, quotes, “if the group were…” and predictions. It’s generated by an AI model (Claude, by Anthropic) and only runs if you tick the box and buy me a coffee (3 €), because every AI analysis costs me money. To do it, the AI reads the chat text, which is prepared in your browser before anything is sent:

• Names are replaced with pseudonyms (P1, P2…), including when they show up inside messages. The mapping back to real names never leaves your browser.
• Phone numbers, email addresses and links are removed. Deleted messages are skipped.
• No photos, videos, voice notes or documents are sent: just a “<media>” marker where one used to be.
• For huge chats the text is capped at 120,000 characters: whole days are kept, spread across the entire history, plus the busiest days. Most group chats fit in full.
• We also send a few short snippets (at most 36, of up to 10 messages each) to measure topics and emotions, plus aggregated stats using pseudonyms (messages per period, busiest days, who replies to whom) so the AI never has to make up a number.
• Every quote we show is checked against the chat: it is always a real message, never something the AI made up.
• The emotions and relationships reading is just for fun: it describes the tone of the conversations, makes no psychological diagnosis and doesn’t assume romantic relationships.
• Responsibility of the uploader: before turning on the AI you must confirm that you upload the chat under your own responsibility and that you have permission from every member of the group to analyze it with AI. Without that confirmation, the AI analysis does not run.
• Our server doesn’t store the data it receives: it only forwards it to the model and returns the result.
• Anthropic processes it to generate the response, under its commercial terms (anthropic.com/legal/commercial-terms): by default it does not use it to train its models and may keep it temporarily for safety reasons. Anthropic is a US company.
• If someone uses a discount code that makes the analysis free, depending on the code it runs on Claude or on Gemini, by Google (the dialog tells you before anything is sent), under its API terms (ai.google.dev/gemini-api/terms). On Gemini’s free tier Google may keep the content and use it to improve its products.

You can see the exact text that is sent to the AI. If you don’t turn it on, nothing leaves your browser: topics are estimated with a local method and the story is built from the stats alone.

4. Payment

• The coffee is paid through Stripe, which handles payment data as an independent controller. We never see or store your card details.
• To issue and send you the receipt, Stripe passes the email address you provide to Stack Studio's invoicing system, which sends it to you with Resend (servers in the EU) and does not keep it once sent. The receipt is issued in Holded without any of your data.
• The payment is not linked to the content of your chat and covers a single analysis: no “unlock” is stored.

5. Cookies, ads and usage statistics

• To keep the app free, Google AdSense ads are shown on the guides. Inside the app (the analyzer, the Wrapped and the game) AdSense is not loaded: the ad screen shows our own Stack Studio ad, which uses no cookies, and it goes away when the AI analysis is turned on.
• Google and its partners may use cookies and identifiers to serve and measure ads and, if you agree, to personalise them. Before that, Google shows you its consent message (certified under the IAB TCF standard), where you can accept, reject or choose purposes.
• Ads never receive the content of your chat: there is only an ad space on the page. More on how Google uses data: policies.google.com/technologies/partner-sites.
• A technical preference may also be stored in your browser. If you reject cookies, the app works exactly the same.
• We count, anonymously and without cookies, how many people use the app (page views, Wrappeds generated, games played…) with Umami, open-source software running on our own servers. It only stores counters and coarse data (language, size range, country, device), never anything from the chat content, and it is not shared with anyone. Your IP address is only used at the moment of the visit to work out the country, and it is not stored.
• We also count page views, without cookies, with Vercel Web Analytics, from the same provider that hosts the app. It only receives the page address, where the visit came from, the country and the device and browser type; it does not store your IP or any persistent identifier, and it never receives anything from the chat content.

6. Data controller

• Stack Studio, S.L., Tax ID (NIF) B67344077, Carrer de Còrsega, 236, local interior, 08036 Barcelona. Contact: info@stackstudio.es.
• You can exercise your rights of access, rectification, erasure, objection, restriction and portability by writing to that address, and lodge a complaint with the Spanish Data Protection Agency (aepd.es).
• Providers involved: Vercel Inc. (app hosting and page-view counts; the servers are in Frankfurt, EU, and the company is US-based, with standard contractual clauses and the EU-US Data Privacy Framework), Stripe (payment), Holded (invoicing, none of your data), Resend (sending the receipt), Anthropic (Claude, AI analysis) and Google (AdSense and, with some discount codes, Gemini). On our own servers (Coolify) we run Umami, to measure usage, the encrypted copy of the share links, the discount codes and a log of AI analyses (date, language, whether it was paid or used a code, and whether it worked; never the chat’s content).

7. What we don’t do

• We don’t connect to your WhatsApp account or its API.
• We don’t try to identify anyone beyond the name that appears in the file.
• We make no psychological diagnoses and don’t infer sensitive information.