Messages AI

Terms of use and privacy policy

Messages AI – Terms of use and privacy policy

Version 16, valid from 11 October 2026. Please read them carefully: by accepting them you confirm that you understand what the app does and that you use it at your own responsibility.

1. Who provides the app

Messages AI is provided by Tilen Matjašič, a private individual (the “author”). Contact: tilenmatjasic6389@outlook.com.

2. What the app does

• With your permission it reads notifications and, through Accessibility, the screens of the messaging apps you turn on in the app.
• It sends the text of conversations, notes and memories to the AI provider you choose (for example Anthropic, OpenAI, Google, Groq, OpenRouter, Mistral) or processes it with a model on your phone.
• It prepares replies in your name. By default you send each one yourself with one tap. Automatic sending sends replies without your review when enabled by you or an installed sending module. By default, each person once receives a short note in their language that an AI assistant helps you write your replies. You can omit this automatic footer with the optional No AI footer module.
• It stores conversations, memories, settings and API keys on your phone.
• If you enable Pro edit learning, the app compares an AI draft with the message you finally send and stores only local structured style signals (for example length, punctuation, capitalization and emoji), not a separate copy of either full text. You can disable and delete this learning.
• If you explicitly choose Pro reply feedback, the app stores only the selected category, local conversation key and time. It does not store a copy of the reply text in feedback history. You can disable feedback use and delete one or all feedback events.
• If you choose a Pro relationship preset for a person, the choice and any custom relationship context are stored locally and used only as a lower-priority tone hint. Your rules, notes, actual conversation and learned style take priority. You can reset it; after a downgrade it remains visible and resettable but is not used for replies.
• To use it you sign in with an account (Google).

2a. Free, Pro trial and subscriptions

The permanent Free plan supports basic app features with a compatible AI method selected by you, including your own API key, a local/on-device model, your local server or a supported free provider. Pro unlocks advanced application personalization and productivity; it does not include paid AI inference funded by Messages AI. Eligible new Google Play customers may receive a seven-day Pro trial. Google Play determines eligibility and shows the actual localized monthly or annual price, renewal period, tax and trial terms before purchase. Builds obtained elsewhere use server-verified PayPal subscriptions where available; a PayPal trial is not offered unless explicitly shown by PayPal. Existing valid VIP purchases are grandfathered as Pro. Cancellation stops future renewal and access continues until the verified end of the paid or trial period. Trial expiry or downgrade does not delete learned data. One account may be active on one device at a time. The developer account has permanent, non-purchasable internal access.
To verify access and prevent payment/trial abuse, the server stores the account ID, current device session, legacy count of previously authorized hosted messages, payment provider, product/base-plan/offer identifiers, a provider subscription reference, purchase/trial state and server-verified expiry. The historical 100-message counter is retained for old app compatibility but does not limit current BYOK, local, local-server or supported free-provider requests. A minimal server-only trial claim may store claim/conversion timestamps and a hashed purchase reference. Billing records contain no chat text, contact names, AI-provider keys or private notes. Google Play or PayPal processes payment details and eligibility under its own terms and privacy policy; the author does not receive full card or bank details.

3. Your responsibility

You use the app at your own risk. You alone are responsible for:
• every message sent from your phone or your accounts, including messages written by the AI;
• choosing the apps and the people the app may reply to, checking its settings and replies and switching it off when needed;
• complying with the law, including the privacy of the people you chat with – their messages are processed by the app and may be sent to the AI provider you chose – and telling them about it where the law requires it;
• the terms of the messaging apps (automated messages may break them and your account may be restricted or banned) and of the AI providers, including their costs;
• not using the app for anything illegal, deceptive or harmful: fraud, scams, spam, harassment, threats, pretending to be someone else, collecting other people's data, or romantic or sexual contact with minors.
The app is meant only for adults (18 or older).

3a. Your choices

The app is a tool. You alone decide and set up: which apps and people it may answer, whether replies wait for your confirmation or are sent automatically, which AI provider and model it uses (in the cloud or on your phone), which API key, rules, descriptions and instructions it follows. The author does not choose, control, review or monitor the providers, the models, your settings or the messages. The consequences of these choices – the content of messages, what is sent to the provider you chose, its costs and compliance with the terms of the services you use – are your responsibility.

3b. Third-party services

AI providers, model hosting (Hugging Face), messaging apps, Google services and PayPal are independent third parties. Their own terms and privacy policies apply. The author is not responsible for their availability, prices, outputs, changes or how they handle data.

3c. Modules

Modules are optional add-ons made by other people (styles, rules, topics, instructions for the AI, support for more apps and small scripts). You decide which modules to install, switch on or remove. A module is data that the app reads: its scripts run in a closed sandbox without access to your phone, files, contacts or the internet. Modules can configure automatic drafting and sending, reply scope, local length/rate limits, protected AI processing, protective rules and the automatic AI footer while enabled. These overrides do not rewrite your saved preferences. Modules cannot change account credentials, stored consent or disclosure history, AI provider keys, chat retention, diagnostics, SMS or personal profile/notes.
"Verified" means the module was signed by the author of the app after a review; it is not a guarantee. The author of a module is responsible for its content and for having the rights to it, and the module's licence applies. The author of the app may remove a module from the community list at any time. Report a module that breaks the law or these terms to the contact above.

4. AI can make mistakes

AI replies can be wrong, inappropriate, out of context or can present you in a way you don't want. The app has rules and checks (for example, it never denies that an AI helps), but none of them is perfect. Protected AI processing hides numbers and other private details before anything is sent to the AI, but it can't recognise every kind of private information.

5. No warranty, limitation of liability

The free trial and any paid access are provided “as is” and “as available”, without any warranty – in particular not that the app is correct, always available, free of errors or suitable for a particular purpose.
To the fullest extent permitted by law, the author is not liable for any direct or indirect damage arising from the use of the app or from being unable to use it, in particular for: messages that were or weren't sent, their content and consequences; damage to relationships or reputation; lost, blocked or restricted accounts; lost data; costs charged by AI providers or other services; and damage caused by services of third parties (AI providers, messaging apps, Google, PayPal).
If a third party makes a claim against the author because of messages sent from your phone or your breach of these terms, you are responsible for it; if you are a consumer, only to the extent permitted by mandatory law.
Nothing in these terms limits liability that can't be excluded or limited under mandatory law (for example for damage caused intentionally or through gross negligence, or for injury to life, body or health), nor your mandatory rights as a consumer.

6. Privacy policy

Controller for data the author processes: Tilen Matjašič, tilenmatjasic6389@outlook.com.
On your phone. Conversations, notes, approved/pending/rejected AI inferences about people, relationship presets, settings, API keys and the log are stored only on your phone, in the app's private storage. The author has no access to them. You can inspect, edit, approve, reject, pause or delete person-profile inferences in the app. They are deleted when you clear the app's data or uninstall it. A backup file is created only when you export one yourself; it is encrypted with a password you choose.
Developer mode. Only the account tilen.matjasic@gmail.com can open developer mode. The developer explicitly chooses an installed app and may start an interaction-recording session. While that selected app is open, Messages AI records accessibility events (such as clicks, long presses, focus and text-field changes, selections, scrolling and window changes) and captures changed stable screen states. Captures can contain a screenshot, the visible accessibility tree and structured controls, crops of controls and labelled/unlabelled icons, on-device OCR, inferred functions and screen regions, a transition graph, labels and notes, and technical app/device data (including version, signing-certificate fingerprint, requested permissions, activity/window information, screen size, density and orientation). Typed event text is not recorded; password node text and pixels identified as password controls are redacted before storage and OCR. Android does not expose the other app's source code or every physical touch coordinate. Nothing is uploaded automatically; the private local data leaves the app only when the developer exports a ZIP through Android's document picker. The developer can stop recording at any time and is responsible for using test content where possible, keeping the ZIP private and deleting captures when no longer needed.
Protection on the phone. API keys are encrypted with a key kept in the phone's secure hardware (Android Keystore). You decide how long chats are kept (People → Storage); "Delete all data on this phone" removes everything the app stored.
Support. "Copy log" puts a technical report on your clipboard only when you tap it; by default it contains no names and no message text. If you send it to the author for help, the author uses it only to solve your problem and deletes it at the latest six months after the problem is closed.
Modules. Installed modules and their settings stay on your phone. When you open the community module list, the app downloads it and the modules you choose from GitHub (GitHub, Inc., USA), which sees your IP address; nothing else is sent. Modules can't send data anywhere.
Reports of AI replies. When you tap Report on a reply written by AI, only the text you send (you can edit it first), the reason, the AI provider and model and the app version are stored in Google Cloud Firestore in the EU (multi-region eur3); no account, names or conversation are attached. The author uses reports to make replies safer and to meet Google Play's rules for AI-generated content. Legal basis: legitimate interest in the safety of the app (Article 6(1)(f) GDPR). Reports are deleted automatically after six months.
AI providers. What is needed for a reply (conversation, notes, approved person-profile inferences, an active relationship tone hint and your description) is sent directly from your phone to the AI provider you set up, under that provider's terms and privacy policy; pending/rejected inferences and inactive paid relationship hints are not sent. The author doesn't receive it. With protected AI processing (on by default) numbers, passwords, codes, phone numbers, IBANs, card numbers, e-mail addresses and links are replaced by placeholders before sending, and screenshots aren't sent to the AI. With a model on the phone nothing leaves the phone.
Which AI providers. You choose the provider and enter your own API key, so the contract for that processing is between you and the provider. Built-in choices include Anthropic (USA), OpenAI (USA), Google Gemini (USA), Groq (USA), OpenRouter (USA, forwards to the model's provider), Mistral (France) and DeepSeek (China – the EU has no adequacy decision for China, so the app warns you before you choose it), or a server you run yourself. Free plans of some providers may use your content to improve their models.
Models on the phone are downloaded from Hugging Face (huggingface.co), which sees your IP address and which file you download; after that, nothing leaves the phone.
Sign-in (required to use the app) with your Google account, or another provider the app offers (Microsoft, Apple, Facebook or X). Google Firebase Authentication (Google Ireland Ltd. and Google LLC) stores your account's name, e-mail address, profile photo, a user ID and sign-in times. When you sign in with Microsoft, Apple, Facebook (Meta) or X, that company also processes the sign-in under its own privacy policy. Legal basis: providing the sign-in you asked for (Article 6(1)(b) GDPR). The data is kept until you delete the account in the app (Account → Delete account) or ask the author to delete it. So that an account is signed in on only one device at a time, Google Cloud Firestore also stores a random ID of the signed-in device, its model and the time of sign-in; it's deleted when you sign out or delete the account. Transfers to the USA are covered by Google's safeguards (EU–US Data Privacy Framework, standard contractual clauses).
Where account data is stored. The session record (device ID, model, time) is stored in Cloud Firestore in the EU (multi-region eur3); Firebase Authentication is a global Google service and may process sign-in data outside the EU, including the USA.
Text recognition and language detection run on the phone with Google ML Kit; Google Play services may download the models and, under Google's privacy policy, send Google technical information about their use.
No ads, analytics or tracking. Donations go through PayPal under PayPal's terms.
Permissions: Accessibility (reading and operating chat screens), notification access (new messages), notifications, optionally SMS and contacts (learning from SMS; not in the Google Play version), screenshots of parts of the screen (text recognition and learning “seen” – kept on the phone as tiny thumbnails; full labelled captures only when the developer explicitly uses developer mode).
Your rights: access, rectification, erasure, restriction of processing, data portability, objection and withdrawal of consent – write to the author. You can export your data yourself at any time (Backup → Export my data). You can lodge a complaint with the Information Commissioner of the Republic of Slovenia (www.ip-rs.si) or with the data protection authority of your country.

7. Donations

Donations are voluntary, give no rights to features or support and are not refunded, except where the law requires it.

8. Changes

The author may change these terms for valid reasons (new features, changes in law or in third-party services). Except where a change is required by law or urgent for security, you are informed at least 30 days in advance in the app. The app then shows the new version and asks you to accept it before it writes any more replies; if you don't accept it, you can stop using the app and delete your data at any time.

9. Applicable law

Slovenian law applies and the courts in Slovenia have jurisdiction; as a consumer you keep the protection of the mandatory law of the country where you live.