01Why this Notice exists
KaleWay is built around a character named KaleAI. KaleAI is not a person — she is an AI assistant powered by Claude, a large language model built by Anthropic, PBC, a U.S.-based AI company.
Because KaleAI is an AI from a third-party provider, when you talk to her, the words you type are processed by Anthropic's servers. We have to tell you this clearly, you have to agree to it before it happens, and you can change your mind whenever you want. This Notice explains all of that.
This Notice is part of our broader Privacy Policy but goes deeper on the AI-specific parts. Where this Notice and the Privacy Policy say different things about AI, this Notice controls.
02What KaleAI is — and isn't
KaleAI is an AI character. She has a name, a face, a voice, and a personality designed by us. The intelligence behind her — her ability to read what you write and respond meaningfully — comes from Anthropic's Claude model.
- A software character we designed and named
- An interface to Anthropic's Claude AI model
- Constrained by instructions we wrote
- Available inside the KaleWay app only
- A real person
- A licensed therapist, dietitian, or doctor
- A medical service
- A way to talk privately to KaleWay's team
When you message KaleAI, you are not talking to us — you are interacting with software that we built on top of Anthropic's AI. We do not read your conversations as they happen, and we do not have a team of humans monitoring KaleAI chats.
03What gets sent to Anthropic
When KaleAI responds to you, the following information is sent to Anthropic so the model can generate her response.
What goes to Anthropic
Sent with each request so KaleAI can respond meaningfully. This is the exhaustive list, verified against the production Edge Function code.
- The text of the message you just typed.
- The last 10 messages in your conversation with KaleAI (so she can respond in context).
- Your first name only — never your last name, email, password, or full name.
- Your weight loss goal (e.g., “lose weight,” “maintain,” “build muscle” if applicable).
- Your daily calorie target (calculated server-side from your profile + latest weight log).
- Your dietary restrictions — the values you set in your profile, sent verbatim (e.g., vegetarian, vegan, gluten-free, dairy-free).
- Your food allergies — FDA Big-9 selections plus any free-text “Other” entries. Allergies travel with every KaleAI request because they are zero-tolerance medical constraints.
- Your cuisine preferences — the cuisines you selected during onboarding or in Profile.
- Your country code (ISO 3166: US, BR, PT, GB, DE, ES, FR, IT, MX, IN, AU, CA, IE, NZ) — drives grocery store and unit references.
- Your health conditions (e.g., pregnancy, PCOS, hypothyroidism) if you've shared them. These are sent verbatim because KaleAI must adjust meal plans and guidance for your situation. They are the most sensitive data we share, and we want you to know that explicitly.
- Your activity level (e.g., sedentary, lightly active, moderately active, active) — used by KaleAI to adjust suggestions.
- Your trigger profile scores — eight numeric scores (0–4) for stress, emotional eating, social pressure, family environment, travel & eating out, holidays, mental barriers, and slip-up spiral.
- Your My Why List — the personal motivations you wrote (or that KaleAI suggested and you accepted), sent verbatim. These are often deeply personal; we want you to know they leave our backend.
- Today's meals from your weekly plan — only the meal name, type (breakfast / lunch / dinner / snack), and calorie count.
- Your latest weight reading — a single number. See subsection 3.1 below for the full picture on weight.
- System instructions written by KaleWay engineers that govern KaleAI's behavior. We do not send instructions specific to your account; the same instructions go to every user's request.
What stays out
These are deliberately excluded from every request.
- Your email address, password, or account session tokens
- Your payment information, subscription identifiers, or RevenueCat data
- Apple HealthKit raw streams — steps, active energy, full weight history beyond your latest reading. Your latest weight reading is a special case explained in subsection 3.1 below.
- Your precise location — we use country-level only (see field 9 in the “Sent” list)
- Photos, audio, or video — KaleAI does not currently support these
- Your full name — only your first name is sent, so KaleAI can address you naturally
- Other users' data — your conversations are isolated to your account; Anthropic only ever receives your conversation
- The names of your Trap Cards, Mind Moments completions, or Behavior Wins — we send a numerical trigger profile, not a list of your titles
- PostHog and Sentry telemetry — analytics events and crash reports flow to those processors directly, not through Anthropic
A note on how detail is sent. Where this list mentions data being sent “as written” or “verbatim,” we mean it: when you set your dietary restriction to “lactose intolerant” or your health condition to “PCOS,” that exact text travels to Anthropic with each relevant request. We do not abstract or anonymize these values before sending. We are explicit about this because it matters — sending raw values is more useful for personalization but also more sensitive. The data is processed under Anthropic's commercial API terms, deleted on the schedule described in Section 4, and never used to train AI models.
3.1 — About your weight specifically
When you ask KaleAI a question, your most recent weight reading is included in the context we send to Anthropic. This is a single number (e.g., 70 kg) — not your full weight history, not your step count, and not any other Apple Health data.
Whether you logged that weight by tapping the scale icon in the app or by syncing Apple Health, it travels the same way: through KaleWay's database, then into the API call to Anthropic. Apple HealthKit weight imports are written to your KaleWay weight log; from there, only the latest entry is selected when KaleAI builds context for a request.
To stop your weight from going to Anthropic, you have two options:
- Revoke AI consent in Profile. This turns KaleAI off entirely — no requests of any kind are sent to Anthropic.
- Email support@kaleway.com requesting that weight be excluded from KaleAI context. This disables KaleAI's ability to tailor calorie targets but keeps the rest of KaleAI working.
04How long Anthropic keeps it
KaleWay uses Anthropic's commercial API — the API used by businesses, not the consumer Claude product at claude.ai. Under Anthropic's Standard Commercial Terms (the terms applicable to KaleWay's account):
Standard retention. Anthropic retains the inputs and outputs of API requests for up to 30 days for service-operations purposes (debugging, abuse prevention, pipeline reliability), after which they are deleted from active systems. In practice, observed deletion of KaleWay traffic is approximately 7 days based on our own monitoring of Anthropic's behavior; we state both numbers because the contractual ceiling is what matters legally and the observed behavior is what matters operationally.
The same retention window covers everything sent in a request — your message, recent conversation context, the profile values listed in Section 3 (including your latest weight reading), and the system instructions we send.
Anthropic does not use this data to train their AI models. This is a contractual commitment that applies because we are a commercial API customer. (If you're familiar with the consumer Claude product, the rules there are different — KaleWay does not use the consumer product, so those rules don't apply to you.)
Two exceptions to the standard window
Trust and safety review. If a conversation triggers Anthropic's automated abuse-detection systems, Anthropic may retain the relevant data for up to 2 years to enforce their Usage Policy. KaleWay does not see or use these flagged conversations; they are managed entirely on Anthropic's side. This is true for any app built on Anthropic's API, including KaleWay.
Legal obligations. If Anthropic is legally required to retain data — for example, under a court order or regulatory request — they may keep it longer.
We do not control either of these exceptions. They are part of how Anthropic operates their service. We disclose them so you have the full picture.
KaleWay's own retention is separate
Your conversations with KaleAI are also stored by us on our backend (Supabase) so you can scroll back through your chat history. Our retention is described in our Privacy Policy and lasts as long as your account is active. When you delete your account, we delete our copies; Anthropic's copies follow their own deletion timeline.
Data residency for AI
Anthropic processes API requests in the United States. For users accessing KaleWay from the European Union or the United Kingdom, this means your KaleAI conversation crosses transatlantic boundaries each time you send a message. We rely on the EU–US Data Privacy Framework (effective July 2023) for this transfer. If the Framework is invalidated by a court or regulator, we will switch Anthropic's contract to Standard Contractual Clauses (SCCs) within 30 days and update this Notice.