How TCHAZO Uses AI

Last updated: November 4, 2026

1. Where AI comes in

TCHAZO uses artificial intelligence to:

• analyze the conversation screenshots and text you submit;
• imagine 3 ways the other person might react to the message you’re about to send. These aren’t their words, and they aren’t a prediction;
• suggest a reply in the tone you pick (gentle, funny, or direct), always respectful;
• suggest example messages to adapt, when you ask. TCHAZO never sends them for you;
• spot a manipulative request or signs of distress, so it can suggest a respectful approach or point you to support resources.

The opening demo doesn’t use AI: its examples are written ahead of time.

2. Who provides the AI

Anthropic’s Claude models, used through Anthropic’s API. We don’t build our own model. Free accounts use a lighter model than the subscription, so their responses may be less detailed.

3. What the AI receives

Only what’s needed for the response you ask for: your screenshots after masking (top banner covered, any areas you hid, metadata removed), conversation text after masking, the message you’re about to send when you ask for possible reactions, your goal, the audience and tone you picked, and, if you share it, your relationship with the other person. Never your name or your email address.

This content goes through our server, is sent to Anthropic, and is then deleted from our server. TCHAZO doesn’t keep it. Anthropic processes it outside Canada, including in the United States, and deletes it automatically within 30 days. If Anthropic’s safety systems flag an exchange as breaking its usage policies, it may be kept for up to 2 years. Anthropic isn’t allowed to use your content to train its models.

Nothing is sent without your explicit consent, which we ask for before the first time anything is sent. You can withdraw it anytime in Settings and still use the demo.

4. What the AI can and can’t do

• It interprets. It can’t read minds: every reading is a likely guess, not a certainty.
• It can be wrong, misread a screenshot, miss context or humor, or reflect biases.
• It isn’t a person.
• Possible reactions are imagined by the AI from the conversation. They don’t come from the other person and don’t tell you what they think.
• It never talks about money, prices, budgets, or anyone’s financial situation, even in a work conversation or a negotiation: it helps with tone and timing, not numbers.
• It doesn’t replace a health professional or legal or financial advice.
• Its example messages are suggestions: read them and make them your own. You decide whether to send them, and you’re responsible for them.
• It doesn’t make any decisions for you.

Trust your own judgment: put the ideas in your own words and fit them to the person you know.

5. What we do to keep it safe

• Instructions to the AI: respect and consent, and no manipulation, harassment, or sexually explicit content.
• Our AI provider’s safety filtering, on top of our own instructions.
• A “The AI’s best read, not a sure thing” label on every analysis.
• “Report a problem” on every result, right from the screen. A report includes only the feature and the category, never your content. We review every report within 7 days.

6. Contact us

Questions about AI in TCHAZO: support@tchazo.com. About your data: privacy@tchazo.com.