Calsa uses AI in two places: to read what you ate from a photo or a phrase, and to answer questions in the AI chat. This policy explains how both work, what data they use, and the limits we put on them. It is part of our Terms of Use and should be read with our Privacy Policy and Consumer Health Data Policy.
The short version
- AI suggests, you decide. Nothing from AI goes into your diary until you confirm it.
- Calories come from the Calsa food database, not from AI guesses.
- AI providers process your data only to answer your request. They may not train their models on it, and they delete it within 30 days, with the narrow exceptions set out below.
- The AI chat is AI, not a person, and it gives no medical advice.
- You can turn AI off at any time. Logging, search, barcodes and your diary keep working.
Logging food with AI
When you take a photo of a meal or type or say a phrase like “two eggs and toast”, AI suggests which foods are there and roughly how many grams. Calsa then looks up each food in its database and shows you a draft. You check the foods and grams and change anything before you save. The calorie and nutrient numbers come from the database entry, so the same food gives the same number every time.
Photos of labels, recipes, menus or your fridge are read the same way and deleted after 24 hours. A photo of a meal stays with the diary entry until you delete the photo or the entry. We remove location and device details from photos before storing them.
In the Calsa iPhone and Android apps, your phone turns speech into text, and the recording is never sent to us or stored. Only that text is sent, the same as a typed phrase. We don’t create voiceprints. The web app has no microphone button. If you use your keyboard’s dictation, your device maker’s dictation service handles it under its own terms, and Calsa receives only the text.
AI can be wrong about foods and portions, especially for mixed dishes, sauces and oils. Always check the draft. AI photo and phrase logging is limited to 15 times a day. Search and barcode scanning have no limit.
The AI chat
The chat answers questions about food and your diary, for example what to eat with the calories you have left, or what changed this week. Depending on your plan, it can see today’s diary, your last 30 days or your whole diary, along with your goal and weight. It uses values from Apple Health or Health Connect, such as steps or active energy, only after you give separate permission.
The chat:
- tells you before your first question that it is AI and not a person;
- doesn’t diagnose, doesn’t talk about medicines or doses, and tells you to ask your doctor or pharmacist about health questions, every time;
- won’t agree with fasting, “making up” for food, or a goal below Calsa’s safe floor, and can’t change your goal for you;
- can’t add or change diary entries on its own. It shows a card, and you tap to save. It also can’t buy, cancel, change settings or delete anything for you;
- doesn’t write to you first and has no name, avatar or personality. It is not a companion, and it doesn’t offer emotional or mental health support;
- answers questions about Calsa from our help articles, and points you to a person at [email protected] when it can’t help.
Each answer is checked sentence by sentence by an automated safety filter before you see it. If a message suggests a crisis or an eating disorder, the chat shows a support response written by people, with helplines for your country, instead of an AI answer. How this works is set out in our Health & Wellness Notice.
Chat answers can be wrong or incomplete. Use them as suggestions, not as facts or advice.
What the chat remembers
On the Calsa and Calsa Premium plans you can turn on chat memory. It is off until you turn it on, and Calsa Start has no memory.
Memory keeps only short facts you tell the chat about yourself that help it answer about food: foods you like and dislike, cuisines, your daily routine or shifts, who cooks, kitchen equipment, budget and units. Each saved fact appears under the answer as “Saved to memory”, with Undo.
Memory is designed not to keep diagnoses, medicines, pregnancy, eating disorders, mental health, weight or body numbers, feelings, facts about other people, or guesses the AI makes about you. If you ask it to remember something like that, it tells you it doesn’t store it. If something like that is saved by mistake, delete it on “What AI chat remembers”. Allergies and foods you avoid go into your “Don’t eat” list, which you save yourself.
You can see every fact, where it came from and when, on the “What AI chat remembers” screen. You can edit or delete facts one by one, delete them all, or switch memory off. Deleting a chat also deletes the facts that came from it.
Who processes your data for AI
These companies process data for Calsa as our service providers, under written data processing agreements:
| Company | What it does for Calsa | What it receives |
|---|---|---|
| OpenAI | Reads meal, label, recipe, menu and fridge photos and logging phrases. Checks chat answers for safety. Answers in the chat if Anthropic is unavailable. | The photo or text of your request, plus the chat context described above. No name, email, account ID, country or time zone is sent with photos and phrases. |
| Anthropic | Answers your questions in the AI chat. | Your message, the diary context for your plan, your memory facts, and Health values only if you allowed it. |
Our AI providers may not use Calsa data to train their models. They delete it within 30 days, unless the law requires them to keep it longer or they need it to investigate a specific breach of their usage rules. Where a provider offers to keep nothing, we’ve asked it to. When we add or change a provider, or send it new kinds of data, we will ask for your permission again before anything is sent. The current list is always on the “Who gets your data” screen in the app.
Calsa itself does not use your diary, photos, chat or memory to train AI models, and does not sell them. We also don’t use AI to make decisions about you with legal or similar effects. Calsa suggests a calorie goal, a portion and chat answers, and you can always change or ignore them. Details are in our Privacy Policy.
Your choices
- Permission. The first time you use a photo, a phrase or the chat, Calsa asks for your permission to send data to an AI provider. Photo and phrase logging and the chat can be allowed separately.
- Turn AI off. Settings has one switch that stops all AI. The chat and AI suggestions disappear, and the next request isn’t sent. Search, barcodes, quick numbers, repeats, your own foods and your diary all keep working.
- Health values. Values from Apple Health or Health Connect go to the chat only if you allow it separately.
- Report an answer. Every answer has “Report this answer”. We review reports within a day. A report includes that answer, your question and the facts used for it, but not the rest of your diary.
Labels and transparency
Every chat answer carries a machine-readable label marking it as AI-generated, in the app and in your data export. When an answer is read aloud, it is labelled “Voice by AI”. We will meet the transparency rules that apply to AI systems in the markets where Calsa is offered, including Article 50 of the EU AI Act.
Contact
Questions about AI in Calsa: [email protected]. Calsa is operated by Genio Group, Inc.