AI Features Notice

Version: 1.3 · Valid from: 3 November 2026

This notice supports the transparency duties of Article 50 of the EU AI Act (Regulation (EU) 2024/1689; applicable since 2 August 2026), Google's disclosure rules for Google Workspace data, and the Terms. It is information, not additional contract terms; the binding rules are in the Terms, the Acceptable Use Policy and the Privacy Policy.

Mailistix offers optional features that use artificial intelligence ("AI Features"). This notice explains in plain language what they do, what data they use, where it is processed, and what you should keep in mind.

1. What the AI Features do

Draft and rewrite requests also contain your name, your email address and your time zone, so that the draft can address people correctly and understand "today" or "tomorrow". Ask AI and meeting briefings contain today's date and your time zone for the same reason.

Feature	What it does	What is sent for processing
Reply and new-email drafts	Suggests a draft that you can edit	The email you answer and the newest messages of its thread (up to 15; senders, recipients, dates, subject, text, attachment names — not the attachments) or, for a new email, the recipient's address; your notes or instructions; the draft you rejected when you ask for a new one; up to five of your own earlier emails with the messages they answered (so the draft sounds like you) or, when no suitable examples exist, a description of your writing style; preferences the app learned from your earlier edits and, until they are learned, up to two recent pairs of an AI draft and the text you sent to that contact; and language or formality preferences you set for the contact. If the email you answer asks for a meeting time and a calendar was read: your free and busy times during working hours for the next seven working days, without titles or participants
Rewrite	Rewrites a draft as you ask	Your current draft, your instruction and notes, and the email you answer with its thread (or the recipient's address, for a new email); for a more formal or friendlier version, also one short example of your own writing
Email and thread summaries	Summarizes an email or a conversation in your interface language	The email (subject, sender, text, links) or the thread (up to 80 messages within a fixed size limit: sender, date and text of each message); when a thread summary is updated, also the previous summary
Ask AI	Answers questions about your mailbox in a chat	Your question and, for a follow-up question, your previous question and the answer; the names and addresses of the people you correspond with most; the languages of your mailbox; and the emails and attachment texts the app found relevant (first short excerpts to choose sources, then the text of the selected emails). To find them, search terms written by the AI are also sent as searches to your mailbox provider (Google or Microsoft); a mailbox connected through IMAP is searched only on your computer
Meeting briefings	Prepares a short briefing for an upcoming meeting	The meeting details (title, time, location, organizer, participants and their responses, invitation text) and up to 14 related emails from the last 120 days (size-limited), plus attachment names
Writing style	Describes how you write when a draft finds no suitable examples	Up to 100 emails you sent recently (shortened), some of them with their subject lines. The description is kept on your device and sent with later drafts
Learning from your edits	After you send a reply that the AI drafted and you changed, works out which preferences your changes show (for example "more concise" or "mention the next step"; never your greeting, sign-off, name or form of address), so later drafts need fewer edits	The email you answered, the AI draft and the text you sent. This is sent once; until the preferences have been worked out, the pair is also included in your next drafts to that contact (see drafts above)

You choose when to use an AI Feature. Mailistix never sends an email, accepts an invitation or takes any other action in your mailbox on its own on the basis of AI Output. A draft stays a draft until you send it.

2. You are interacting with AI

Drafts, rewrites, summaries, briefings and Ask AI answers are generated by AI models. In the app they are marked as AI-generated: a finished draft shows "AI draft", summaries are titled "AI summary" or "AI thread summary", meeting briefings "AI briefing", and every Ask AI conversation starts with the note that you are talking to an AI and that its answers can be wrong. Ask AI is an AI system, not a person.

3. Limits of AI

AI Output can be wrong, incomplete, out of date, or inappropriate in tone, even when it sounds confident. It may misread an email, invent facts, names, dates or commitments, or miss important points. Please:

read and check every draft before you send it — you are the sender and are responsible for what you send;
check important facts in the original email, especially amounts, dates, deadlines and legal or contractual statements;
do not rely on AI Output for medical, legal, financial or other professional decisions.

The AI Features are off until you enable them in the AI consent dialog. You can switch them off at any time in Settings › Privacy & AI; after that, no data is sent to AI models until you enable them again.

5. Where your data is processed

Your request goes from the app over an encrypted connection to our server functions, hosted by Supabase in the EU (in Frankfurt, Germany, by default; on Amazon Web Services).
From there it goes to one of our AI model providers: Microsoft Azure OpenAI (EU Data Zone; currently Sweden Central) or Google Cloud Vertex AI (Gemini models).
If you told us you are based in Europe (the EU, the EEA, the United Kingdom or Switzerland), or have not answered the region question, all AI processing takes place in the EU: Azure's EU Data Zone and Vertex AI's EU multi-region.
If you told us you are based outside Europe, most requests to Gemini models run on Google's global endpoint, which may process data in any country where Google operates. Some steps of Ask AI still run in Vertex AI's EU multi-region, and requests to Microsoft always run in its EU Data Zone. You can change your region at any time in Settings › Privacy & AI.

The current list of providers is in our Subprocessors list.

6. What happens to your data afterwards

On our servers: we do not store the content of your requests or of the AI Output. The technical logs of our server functions contain no content; our hosting provider keeps them for up to 90 days. We store usage metadata (which AI Feature, model, token counts, Credits, time, success or error) for billing, cost control and security, for 180 days. If our safety check holds back an answer, we record that this happened, without the answer text, for security purposes.
At our AI providers: the providers process the data only to answer the request, under data processing terms that prohibit using it to train their models. Microsoft is called with storage of the request switched off. The providers may keep data for a limited time for operational and abuse-monitoring purposes: Google Vertex AI may hold inputs and outputs in memory for up to 24 hours (caching) and may log prompts that its automated systems flag as possible abuse for up to 90 days, where authorized Google employees may assess them; Microsoft Azure OpenAI may store prompts and outputs for up to 30 days to monitor for abuse, where authorized Microsoft personnel in the European Economic Area may review them.
On your device: the app keeps drafts, summaries, briefings, your Ask AI conversations, the description of your writing style and AI-related preferences locally in its encrypted database. When you send a reply that the AI drafted and you changed it, the app stores the email you answered, the AI version and your sent version on your device (the latest ten per contact, for at most 180 days) and sends them to an AI model to work out your preferences (see the table above). The preferences it learns are stored on your device and sent with your later draft requests, so that drafts better match how you write.
No training: we do not use your emails, prompts or AI Output to train or improve AI models, and our providers may not either. The only "learning" is the per-user personalization described above: it is stored on your device and is used only as part of your own requests; no model is trained or changed.
No human review: our staff do not read your emails, prompts or AI Output, except where you ask us to look at specific content (for example in a support case), where it is necessary for security or abuse investigations, or where the law requires it.

7. Google user data

Mailistix's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements, and the information received from Google Workspace APIs will also adhere to the Google Workspace API User Data and Developer Policy. Data from Gmail and other Google Workspace APIs is not used to develop, improve or train generalized or non-personalized AI or machine-learning models. It is sent to an AI model only after you switched on the AI Features in the AI consent dialog, and only for the features described in section 1.

8. Marking of AI-generated text

Emails that Mailistix sends from a draft written with an AI Feature carry a machine-readable marker in the message header, X-AI-Generated: Mailistix, which does not change the visible text. This applies to all users, whatever region they chose. A draft counts as written with an AI Feature once an AI draft or rewrite was put into it, even if you changed the text afterwards. The marker is added to every such email sent from a Gmail account or from a mailbox connected through IMAP. For Outlook and Microsoft 365 accounts it is added to new emails; replies and forwards normally go through Microsoft's own reply and forward functions, which do not accept such headers, and then carry no marker. A draft that you send from another app (for example from Gmail or Outlook in the browser) carries no marker either. AI Output inside the app is labeled as AI-generated (section 2). Our AI providers may in addition embed an invisible watermark in the text their models generate; whether they do is their decision, and Mailistix neither adds nor checks such watermarks.

9. If you use the AI Features for your business

Mailistix is the provider of the AI Features. As a Business Customer, you use them under your own authority and are therefore a "deployer" under the EU AI Act (Article 3(4)). In particular:

AI literacy (Article 4). Make sure that the people who use the AI Features for you understand, to a sufficient degree, how they work and what their limits are. Sections 2, 3 and 8 of this notice and the labels in the app can help.
Published text (Article 50(4)). If you publish text generated or changed with the AI Features to inform the public on matters of public interest, you must disclose that it was generated or manipulated by AI, unless a person reviewed it and you or your organization hold editorial responsibility for it.
Intended purpose. The AI Features are intended for reading, summarizing and writing ordinary business correspondence. They are not intended or designed for the high-risk uses listed in Annex III of the EU AI Act, for example filtering or evaluating job applications, evaluating employees, or assessing creditworthiness, and the Acceptable Use Policy does not allow such uses.
Your correspondents. If you process the emails of other people with the AI Features, you are responsible for having a legal basis and for informing them as data protection law requires (Section 9.6 of the Terms and the Data Processing Agreement).

10. Questions

Contact legal@mailistix.com. More details are in the Privacy Policy.