Transparency
AI Disclosure
Caliber’s coaching is generated by AI. This page explains, in plain language, what that means — what provider we use, what gets sent, what we don’t keep, and how to turn it off.
What the AI does
You screenshot a conversation. Caliber reads the text from that screenshot on your device, then sends that text to an AI provider to generate three reply suggestions with the reasoning behind each. The AI also helps maintain your Playbook — cross-conversation context so the coaching stays relevant over time.
Which AI provider
Caliber uses Anthropic (the Claude API) to generate coaching responses. When you use the coaching features, the conversation text read from your screenshot is sent to Anthropic for processing and the suggestions are returned to you.
What we send — and what we don’t keep
- We send Anthropic the conversation text read from your screenshot, only to generate your reply suggestions.
- We do not store your screenshots or conversation text on our servers. We retain only analysis metadata — momentum scores and the Playbook context.
- Your content is not used to train AI models. It is processed only to respond to your request, subject to Anthropic’s Commercial Terms of Service.
Your consent, and how to turn it off
Before any data leaves your device for AI processing, Caliber asks for your explicit consent and names Anthropic by name. You can review and revoke that consent at any time in Settings → AI Provider Consent. If you turn it off, AI coaching features stop working until you turn it back on. If we ever change AI providers, we’ll ask for your consent again before sending any data to a new provider.
What the AI is not
AI-generated coaching can be inaccurate, incomplete, or wrong for your specific situation — apply your own judgment. Caliber is coaching software, not therapy, medical advice, or professional counseling, and it does not diagnose or treat any condition.
If you’re experiencing distress or a crisis:
United States — call or text 988 (Suicide & Crisis Lifeline), 988lifeline.org.
Outside the U.S. — find local help at findahelpline.com.
Reporting a bad response
In any AI conversation, long-press the response and tap Report, or email [email protected]. We review every report within 72 hours.
For the full picture, see our Privacy Policy and Terms of Service.