Generative AI Notice
Contents
- Posted: 2026-10-15 (first posted 2026-10-06)
- This notice tells you in advance that Aria includes features that operate on generative artificial intelligence, and explains their output. It is provided under Article 31 of Korea’s Framework Act on the Development of Artificial Intelligence and the Establishment of Trust.
1. Aria operates on generative AI
Aria’s AI features run on generative AI models developed by other companies.
| Provider | Where it is used | What it produces |
|---|---|---|
| Anthropic, PBC — Claude models (currently Claude Sonnet 5.5) | Chat, daily and weekly briefings, automatic organizing of tasks and notes, Synapse (knowledge graph) concept extraction, AI memory, weekly reflection questions, recording summaries, web look-ups (Anthropic web search — finding up-to-date information in chats and “Look it up” on tasks) | Text (replies, summaries, titles, classifications, tags, questions) |
| OpenAI OpCo, LLC — embedding model | Semantic search | Numeric search vectors (no text shown to you) |
We may change the model or version on our server for quality and reliability. If the provider changes, we will update this notice and our Privacy Policy.
2. How AI output is shown
- AI replies appear in chat as messages from the assistant “Aria”, and briefings appear as briefing cards, visually distinct from text you write.
- When AI organizes tasks or notes, or creates or changes events, the result is shown, for example as a result card, so you can review it.
- If you copy or share AI output outside the app, we recommend saying that it was written by AI so that recipients are not misled.
3. AI is at the core of the Service and works only after you agree
- Aria is a personal assistant built on generative AI that understands your tasks, calendar and notes to give you tailored briefings, memory and connected notes, so AI processing is an essential part of the Service. For that reason the app has no setting that stops AI features on their own while you keep using the rest of the app.
- Before anything is sent to an AI provider, the app explains what information is sent to which provider and when (including the destination country, the retention period and what happens if you decline) and asks for your explicit consent. New users consent in the AI step of initial setup; existing users who have not yet consented (including those who had AI features off in an earlier version) see a consent screen the next time they open the app. Nothing is sent to an AI provider before you consent.
- People under 14 may not use the Service.
- You may decline, but if you do, you cannot use Aria. Without consenting, you can still go from the consent screen to Settings to manage or delete your account.
- You can withdraw your consent at any time. In app version 1.1 or later, use Settings → AI → “Withdraw AI Consent” to withdraw while keeping your account and records; in earlier versions, email our privacy officer. After withdrawing, you cannot use Aria until you consent again. To delete your account and data as well, choose “Delete account” or “Delete all my data” in Settings → Account; this also ends transfers to AI providers (Privacy Policy, Sections 1.2 and 10).
4. Limits of AI output
- AI output is generated probabilistically and may be wrong, out of date, incomplete or biased.
- Answers based on what was found on the web may be wrong or out of date depending on those web pages; check important information on the source page.
- Always confirm medical, legal, tax, financial and safety matters with a qualified professional. AI output is not professional advice.
- In an emergency, do not rely on AI; contact your local emergency services immediately.
- Aria is a personal productivity service that does not make decisions with a significant impact on people’s lives, safety or rights, such as hiring, credit or medical diagnosis. We operate it as a service that is not high-impact AI under the Act.
5. How your data is handled
- When you use AI features, the information needed for the request (the message you send, photos or a current location you attach (area name only — from app version 1.1, no coordinates), related tasks, notes, events, memories, etc.) is sent to the AI provider through our server. Your account identifier and email are not sent.
- When you use web look-ups, the search query the AI writes passes through Anthropic to the search provider Anthropic uses (Privacy Policy, Section 1.2).
- Neither we nor the AI providers use your content to train AI models.
- Everything is sent over encrypted connections (TLS), and our server uses per-account access control (row-level security) so that other users cannot see your data. When you delete your account, your data on our server is deleted too (Privacy Policy, Sections 2 and 7).
- What is sent, international transfers and retention periods are described in Sections 1 and 5 of our Privacy Policy.
6. Reporting inappropriate AI replies
If the AI gives an inappropriate or harmful reply, touch and hold the reply in the chat (on Mac, right-click) and choose “Report Inappropriate Response”. An email opens, pre-filled with the reply time, app version and reply text, which you can review or edit before sending. You can also write to contact@lipson.kr directly (with a screenshot if possible). We will review it and take appropriate action.
7. Provider
Lipson Inc. · Representative KIM HYUNGI · 3F, 16 Eonju-ro 147-gil, Gangnam-gu, Seoul 06048, Republic of Korea · contact@lipson.kr