Legal

AI in Chirpy

Last updated 17 August 2026.

In short.Chirpy's assistant is an AI, not a person — you are never talking to a human here. It runs on models made by Anthropic, and to answer you it sends them the parts of your household's data your question touches. It only ever runs when you use an AI feature. It is off by default for children, can be switched off per person, and your content is never used to train anyone's model. It can be confidently wrong, so check anything that matters.

1. Who provides it

Every AI feature in Chirpy is powered by Anthropic(United States) — the Claude family of models. We do not run our own models, and no other AI company receives your data. Anthropic processes it on our instructions under a data-processing agreement that forbids using it to train models.

2. You are talking to a machine

Everything the assistant writes is generated by software. It has no person behind it, it does not know you, and nothing it says has been checked by a human before you read it. Where the law requires us to say so explicitly — including the EU AI Act's transparency rules, which have applied since 2 August 2026 — this is that disclosure, and the app says the same thing where you actually use the feature.

3. What each feature sends

Nothing is sent unless you use the feature. Each one sends only what it needs to do its job:

Feature	What leaves our servers
Ask Chirpy	Your question, plus the household context needed to answer it — which can include recent transactions, document titles and dates, lists, meals and records
Scanning a document or receipt	The photo or PDF itself, as an image, plus any text you added
Importing a bank or card statement	The text of the statement file you uploaded, to turn rows into transactions
Gmail receipt import	The sender, subject and body of messages that look like receipts. Processed and discarded — only merchant, amount, date and the message ID are kept
Kitchen and recipes	Recipe details, and the allergies and dietary needs of members who have consented to that
Money and tax helpers	The figures on the screen you are using — your income, filing status and country for tax estimates
Invest insights	Your holdings, contributions and balances for the account you are looking at
Wishlist price lookup	The product web address you pasted, and its title. Anthropic then fetches that shop page itself, from their systems

We keep a record of that a call happened — which feature, which model, how many tokens, what it cost — so credits and spend caps work.We do not keep the text of your prompts or the AI's replies on our servers beyond whatever you choose to save into Chirpy yourself.

4. Health and other sensitive details

Allergies, dietary needs and health records are only sent when you have turned AI features on for yourself, and only the minimum a feature needs — an allergy list so a suggested recipe avoids it, for example. Another member switching AI on does not send your health details. See the health section of the Privacy Policy.

5. Training

Your content is not used to train AI models — not by us, and not by Anthropic, whose commercial terms for the interface we use exclude it. We also do not use it to build advertising profiles, and we do not sell it.

6. Nothing here decides anything about you

Chirpy's AI suggests, drafts, extracts and estimates. It does not make automated decisions that produce legal or similarly significant effects — it does not price anything, approve or refuse anyone, score you, or decide what you may access. A person is always the one who acts on what it produces. That means the rules on solely-automated decision-making (UK/EU GDPR Article 22, and California's automated decision-making technology rules) are not engaged. If that ever changes, we will say so here first and give you the choices those rules require.

7. It gets things wrong

Nothing the AI produces is financial, tax, legal, medical or immigration advice. Check anything important, and never rely on it for a medical safety decision — always verify ingredients yourself rather than trusting an allergy summary.

Language models state wrong things fluently and confidently. Amounts can be misread off a receipt, a date can be invented, a tax figure can be out of date. Treat every output as a draft written by something fast and fallible.

8. Turning it off

Per person — an owner or partner can switch the assistant on or off for any member in Settings → Household.
Children start with it off. An adult has to turn it on deliberately, and can turn it back off.
Just don't use it. Nothing is sent in the background. No AI feature runs on a schedule, and none runs while you are elsewhere in the app.
Gmail import is a separate opt-in and can be disconnected in Settings → Connectors.

9. Fair use

AI runs on a monthly pool of credits per household, and an owner can give an individual member a smaller slice. There is also a ceiling across all of Chirpy so that a bug or an abusive account cannot run up a bill that takes the service down. When a limit is reached the feature stops and says so — nothing is silently degraded.

10. Questions

If you want to know what a specific feature sent, or you want AI switched off for your whole household, contact us at the contact address we publish here before Chirpy opens to the public. The full picture of who receives data is in the Privacy Policy and the subprocessor list.