Legal
Misura AI Transparency Notice
Document: AI Transparency & Disclosure Notice
Document key: ai-transparency-and-disclosure
Responsible entity: Rafael Antonio Berrios Cuneo ("Rafael Antonio Berrios Cuneo", "we", "our" or "Misura")
Address: Tacna, Perú
Contact email: rafoantoni@gmail.com
Website: misura-app.com
Effective date: October 9, 2026 (revision extending sections 2.1 and 8 so that the Misura assistant can look up your intervals.icu activities, and completing the Anthropic row of section 8.1 with weight, training and activities; previous version: October 8, 2026)
Scope: Worldwide (all countries and territories where the Service is available)
Governing law: Peru (without prejudice to any mandatory local rights available to you)
0. What this document is for
0.1. This Notice explains where Misura uses artificial intelligence, what that AI produces, what can go wrong and what you need to do in each case. It is written to be read, not to cover us: if any part is unclear, write to rafoantoni@gmail.com and we will rewrite it.
0.2. This Notice forms part of the Terms of Service and must be read together with the Health & Medical Disclaimer and the Privacy Policy. Where this Notice discusses the accuracy of health data, the Health Disclaimer continues to apply in full.
0.3. THE MOST IMPORTANT THING, IN ONE SENTENCE. Nobody has checked the allergens Misura shows against the package in your hand: they may be incomplete or wrong, and you must always check the product label. If your allergy can cause a severe reaction, do not use Misura as your source for deciding whether a food is safe.
1. What is, and what is not, "AI" in Misura
1.1. When the app says "AI" or "AI-generated", we mean a third-party language model (see section 8) that produces text or numbers from what you give it. That model does not measure, weigh or analyse: it predicts.
1.2. Not everything Misura calculates is AI, and we will not label it as if it were. Specifically:
- The Misura Calculation and your nutrition targets come from published, deterministic formulas (Mifflin-St Jeor, Katch-McArdle, EFSA/AESAN reference values). The same inputs always give the same result. There is no AI in them.
- Food suggestions in search and logging are ranked by a deterministic scoring algorithm over your history and the catalogue. There is no AI in them. (Not to be confused with the recommendations that accompany the daily commentary, which an AI does write — see 2.1.)
- Training metrics (TSS, CTL, ATL, form, ACWR, VDOT, zones) are formulas. There is no AI in them. Their limitations are explained in the Terms of Service.
1.3. Labelling something "AI" when it is not would be false, and it would wear out the labels that matter. That is why this Notice lists the surfaces one by one.
2. Where Misura uses artificial intelligence
2.1. As of the effective date of this Notice, Misura uses AI in the following features. The "what it writes" column is what determines severity.
| Feature | What the AI does | What it writes |
|---|---|---|
| Chat (AI tab) and assistant | Converses, searches the catalogue and proposes amounts; looks up your day, your weight and your training and, if you have training mode turned on, your intervals.icu activities (pace, heart rate, splits, zones) to analyse your performance; opens screens and changes the app's appearance | Puts foods and grams on your plate (you press Log). Opens screens and changes the appearance with an Undo button. It can prepare changes to your diary from the last 30 days (fix, move or delete up to 5 entries), log your weight, or create a food or a recipe of your own, kept private: it never applies them by itself — it shows you a card and they only apply if you confirm it, with Undo. It does not change your training plan |
| Food photo | Identifies foods in an image and estimates weight | Proposes foods and grams for your plate |
| Describe a meal | Turns free text into foods and amounts | Proposes foods and grams for your plate |
| Voice logging | Interprets what you dictate to Siri or your assistant | Writes to your diary after your spoken confirmation, with no review screen |
| Estimated food | Generates the full nutrient profile of a food that is not in the catalogue | Creates a food that is visible to all users |
| Composed dishes | Composes a dish with its ingredients and amounts | Creates a dish that is visible to all users |
| Filled-in micronutrients | Fills micronutrients missing from branded products | Completes data that is visible to all users |
| Allergen classification | Decides which allergens a food contains when rules are not enough | Flags allergens visible to all users |
| Daily coach commentary | Writes the text summarising your day and suggests foods to cover what you are missing | Shows text and food recommendations in the app. It writes nothing to your diary |
| Blog | Writes the articles | Publishes text on the website and in the app |
| External assistant connection | An assistant of yours (for example, claude.ai) can log meals if you connect that integration | Writes to your diary |
2.2. Misura does not run any AI model inside your phone. Everything above happens on servers, with the providers listed in section 8.
2.3. This list can change. The current version is always the one in this document; the effective date shows when it was last revised.
3. Allergens: the one feature where a failure can send you to hospital
3.1. How the data is produced. Misura does not read your product's label. Where the flag comes from depends on the food and is not always the same: for some, it comes from a label a person transcribed (in most of those cases an Open Food Facts contributor, not Misura); for others an automated rule system set it by looking for known ingredients in the name and composition; for others an AI model decided; and for others there is no record of where it came from. None of those routes has seen the package in your hand.
3.1 bis. In recipes the calculation is cruder still. On the recipe screen the allergens do not come from the catalogue: the app infers them by looking for a short list of words in the ingredient names. It has never read an ingredient list. Treat it as what it is: a hint, not a check.
3.2. What can go wrong, specifically.
- An allergen that is present can be missing here.
- An allergen that is not present can appear here.
- A product can change its recipe without changing its name.
- Traces and cross-contamination are never covered.
3.3. What you need to do.
- Read the label before you eat. It is the only source that answers for the product you bought.
- Eating out, ask whoever cooked or served it.
- If your allergy can cause a severe reaction, do not use Misura to decide whether a food is safe. Use it for everything else.
3.4. Scope of the list. Misura's catalogue and your allergy profile handle the fourteen allergens in Annex II of Regulation (EU) No 1169/2011, the same list European labels must declare. Recognising the list is not the same as having checked a specific product. The recipe-screen detection (section 3.1 bis) is narrower still: it searches a short list of words, lumps molluscs together with crustaceans and does not look for lupin at all, so a recipe may fail to show an Annex II allergen that is actually in it.
3.5. Where each flag came from: today the app does NOT tell you. Internally, every allergen in the catalogue carries one of four states: checked by a person, inferred automatically, decided by AI or not checked. That state is not shown on any screen as of the effective date of this Notice. Since you cannot know which of the four you got, the only safe reading is to treat them all the same way: as data that has not been checked. If the app ever shows that provenance, we will say so here and on the screen itself.
3.5 bis. A warning about the word "checked". "Checked by a person" means somebody transcribed a label — in most cases an Open Food Facts contributor — not that Misura verified the product. "Inferred automatically" does not mean an ingredient list was read: for a large share of those rows none was, and their real provenance was lost. None of the four states is a check of the package in front of you.
4. Numbers that go into your diary
4.1. Chat, photo, describe and voice logging pick a food and assign it an amount in grams. Getting the food right and getting the weight right are two separate things, and the system misses on both.
4.2. The worst case is photos: an image does not contain weight. The grams come from comparing with similar plates, and a deep bowl or a generous serving throws that off badly.
4.3. What you need to do.
- Check that the food is the one you ate, not just one that looks like it.
- Check the grams before you log. That is the number that moves your calories and macros for the day.
- If something is wrong, fix it right there, before you log. If you also want to tell us, the channel is in section 7.
4.4. Voice logging is different, and worth knowing about. It is the only path that does not show you a review screen: you confirm out loud and the entry is saved. If you use it, check your diary afterwards.
5. Foods, dishes and micronutrients generated by AI
5.1. Estimated foods. When you search for a food that is not in the catalogue, Misura may generate its full nutrient profile with AI from the name, filling every value with what is typical for similar foods. It does not come from a lab or an official table.
5.2. Composed dishes and filled-in micronutrients. In the same way, the AI can compose a dish's ingredients and amounts, or fill in micronutrients missing from a branded product. Those micronutrients are then compared against dietary reference targets, which can lend them an appearance of authority they do not have.
5.3. This data is public, and you cannot edit it. Foods, dishes and values generated by AI are stored in a shared catalogue and are visible to other users and on the public pages of misura-app.com. One error becomes everybody's error. The app does not let you correct a catalogue entry: the edit pencil only appears on foods you created yourself, and editing one of those creates your own copy — it does not change what anyone else sees. If you see a plainly wrong value, the only route is section 7.
5.4. What you need to do.
- With the package in front of you, the label wins. Its printed values come from the manufacturer; these do not.
- For what you eat every day, look the food up in the catalogue before letting it be estimated.
- If a value is plainly wrong, write to us (section 7). Do not wait for it to be fixed before deciding anything that depends on it.
6. Blog
6.1. Misura's blog articles are written by an artificial-intelligence model and are not reviewed by a person before publication. They do not go through a healthcare professional or a human editor.
6.2. Every article carries that disclosure visibly on the page itself, both on the website and in the app.
6.3. What this means for you. There can be factual errors, wrong figures or advice that does not fit your situation. A general article never knows whether you are pregnant, on medication, or living with a condition that changes everything.
6.4. What you need to do. Treat it as general information, not as an instruction. Before you change how you eat because of something you read on the blog, talk it over with a professional who knows you.
7. How to tell us the AI got something wrong
7.1. The channel is email, and as of the effective date it is the only one. Write to rafoantoni@gmail.com with the name of the food and what is wrong: the food is not that one, the amount is off, the nutrition values are wrong, or the allergens are wrong. The app does not have a report button yet; if we add one, we will say so here. This matches section 14.1 of the Acceptable Use & User Content Policy, which states that email is the only reporting channel enabled.
7.2. What we do with your report, and what we are not promising. We read it and use it to correct the catalogue. There is no priority queue and no committed turnaround, not even for allergens. If your report is about an allergen, do not wait for us to fix it: check the label and stop relying on that entry in the meantime.
7.3. Reports you send us are kept alongside your email address for as long as they are useful for correcting the catalogue. We do not publish your report or your identity.
7.4. If the problem is urgent, say so in the subject line.
8. What data of yours leaves Misura, and where it goes
8.1. For these features to exist, Misura sends data to third-party AI model providers:
| Provider | What is sent | In which features |
|---|---|---|
| Anthropic (Claude models) | The text you write, the transcript of what you dictate, and the food photos you attach; for the daily commentary, the foods you logged today and over recent days together with your targets; and, when you ask the assistant, summaries of your weight and your training and, if you have training mode, the intervals.icu activities you ask about, summarized one by one (pace, heart rate, per-km splits, time in zones, among other data). Never the second-by-second series, the route or any location, nor the activity's title or notes | Chat and assistant, photo, describe, voice, estimated foods, composed dishes, micronutrients, allergens, daily commentary |
| OpenAI | The food text to be searched and the names of the foods you logged over recent days | Semantic food search and the weekly diet-variety indicator |
8.2. A photo of your meal and a description of what you eat are personal data concerning your health, and so are your weight, your training and your activities. Activities are read from intervals.icu by our servers at the moment you ask, using the connection you already have saved, and we do not keep a copy. The processing of that data, its legal basis, retention and your rights are set out in the Privacy Policy.
8.3. If you do not want your data sent to these providers, do not use the AI features, or do not ask the assistant about your weight, your training or your activities: that data is only sent when you ask. The rest of Misura — manual logging, catalogue, calculations, charts — works without them.
8.4. In addition to what goes out to the providers, every action the assistant performs on your account is recorded in our audit log of AI actions, together with the arguments the model requested it with. If the assistant prepares a card to log your weight, that entry stores the weight in kg proposed (and the date, if the model gave one), whether or not you confirm it. That log does not leave Misura, is kept for a maximum of 90 days and is deleted when you delete your account; exactly what it stores and for how long is set out in Sections 6.2 ter and 11.4 ter of the Privacy Policy.
9. Limits we cannot remove
9.1. No disclaimer in this document turns an estimate into a measurement. An AI model can be wrong with apparent confidence, and it is wrong more often with uncommon foods, prepared dishes and regional products.
9.2. Misura is not a medical device and does not diagnose, treat, cure or prevent any medical condition. Consult a healthcare professional for any medical decision, as set out in the Health & Medical Disclaimer.
10. Changes to this Notice
10.1. If we change where or how we use AI, we will update this document and its effective date. Where the change affects what you need to do, we will also announce it inside the app.
11. Canonical language
11.1. The Spanish version of this Notice is the canonical version. Any translation, including this English one, is offered for your convenience. In the event of a discrepancy the Spanish version prevails, to the maximum extent permitted by applicable law and without prejudice to mandatory local rights.
12. Contact
- Entity: Rafael Antonio Berrios Cuneo
- Address: Tacna, Perú
- Email: rafoantoni@gmail.com
- Website: misura-app.com
End of document.