Wie Kahu KI einsetzt

Our principles

• Nothing runs in the background. A model only sees your information when you send a message or ask for a note to be structured. There is no silent analysis of your care record.
• The smallest useful context. Each request carries only what the reply needs, filtered by what your role is allowed to see. We log which entries were read so you can check.
• No training on your content. We do not train models on your data and we do not permit providers to train on it. Your content is used to generate the reply and nothing more.
• People publish the library. Models help our editors draft and illustrate training and wellness content. An editor decides what is published, and your personal information is never involved.
• Not a veterinarian. The Advisor explains and points you to the right next step. It does not diagnose, treat or prescribe, and it is not an emergency service.

Where Kahu uses AI today

• The Advisor. A chat in the app that answers care questions using the published breed encyclopedia, the selected Pet's Meal Plan when you ask about feeding, and a limited, permission-filtered part of that Pet's care record. A model also writes the short title shown in your conversation history and suggests follow-up questions you might ask next.
• Structuring a typed note. When you type something like "gave Jett his flea tablet this morning", a model turns it into a structured entry for you to confirm before anything is saved. The text you typed and your Pet's name and species are sent for this.
• Content Studio (staff only). Our editors use models to draft, review and illustrate the training and wellness library, including automated brand and safety review passes before an editor publishes. Your personal information is not part of this.

What is sent and who processes it

Advisor messages, the encyclopedia text used for a breed answer, the Meal Plan read for a feeding question and the care record entries the Advisor retrieves are sent through Vercel AI Gateway to a model provider. Today those providers are Anthropic, Google and OpenAI. They are listed, with their locations, on the sub-processor list.

We keep a log of which care record entries and Meal Plans each Advisor reply read, so you can see it and we can audit it. Advisor use is metered per Family: 5 messages a month on the free tier and 150 a month on Kahu Premium, as a fair use allowance. Metering is enforced on our servers.

Youth Members cannot use the Advisor. The server refuses the request and the entry point is hidden. You can delete an Advisor conversation from its history at any time; the rest are removed when an account is permanently deleted.

What our AI does not do

• It does not diagnose, treat or prescribe. It is not a veterinarian and it is not an emergency service. For urgent symptoms, use the emergency flow in the app or contact a clinic.
• It does not make decisions about you. Nothing about your account, your Family or your access is decided by a model.
• It does not read what your role cannot see. Permission filtering is applied before context is retrieved, not after.
• It does not run in the background. No message, no request. Closing the app sends nothing.
• It does not score the weekly checkup. The checkup triage tier comes from fixed rules applied to the flags you select, not from a model.
• It does not learn from your dog. Neither we nor our providers train models on your content.

Your choices

If you do not use the Advisor or note structuring, no part of your information is sent to a model provider. Everything else in Kahu works without them.

Questions about how Kahu uses AI go to privacy@kahu.dog. The privacy policy covers how we handle personal information more broadly.