Help · AI + privacy
Where the AI looks, and where it doesn’t.
ConKarma uses AI to generate the monthly family narrative and the coaching drift report, and to screen submitted text for child-safety. People reasonably want to know what it can see. Short answer: the family narrative reads only what you’ve marked as eligible; adult-zone (Ember) content is never used for the family narrative and is processed by AI only under the explicit adult-feature consent you give; and nothing is used as training data.
ConKarma uses AI to generate the monthly family narrative and a coaching drift report, and those features can see only what you've explicitly marked as eligible — never adult-zone (Ember) content, and never as training data. Every journal entry has three visibility levels: personal (only you, the default), cell-shared, or family-narrative-eligible. The AI never reads a personal entry; it only reads what you've deliberately opened for it by opting in to the monthly narrative.
Ember (adult-zone) content is never used for the family narrative or the coaching report — the data layer doesn't expose it to those features. AI touches Ember content only under the explicit adult-feature consent you give (GDPR Art. 9(2)(a)): an AI-powered Ember feature you opt into, or child-safety moderation of text you submit there, which retains only a hashed fingerprint and a verdict. It is never processed for any other purpose, and never as training data. Our AI runs through a vendor-neutral AI Gateway (Anthropic, OpenAI, and Google), and no provider may use customer data for training without an explicit opt-in we have not taken. The narrative is an editable draft, and drift detection is a suggestion to consider support, never a clinical diagnosis.
Only family-narrative-eligible entries
Every journal entry has three visibility levels: personal (only you), cell-shared (your cell sees it), or family-narrative-eligible (the AI may include it if you opt in to the monthly narrative). The default is personal. The AI never reads a personal entry. It only reads what you’ve explicitly opened for it.
Adult-zone AI runs only on your explicit consent
Ember — Spice Calendar, dares, intimate journals, whispers — is never read by the family narrative or the coaching report; that data layer does not expose it to those features. The only AI that ever touches Ember content does so under the explicit adult-feature consent you give (GDPR Art. 9(2)(a)): an AI-powered Ember feature you opt into, or child-safety moderation of text you submit in Ember — both on the same minimised, hashed basis described below. We never process Ember content with AI for any other purpose, and never for training.
No training on your data
Our AI runs through a vendor-neutral AI Gateway, with Anthropic (Claude), OpenAI, and Google as the providers behind it, picked per feature. The contracts we operate under do not allow any of them to use your data to train their models without explicit opt-in. We have not opted in, and will not without telling you first.
You can always edit or delete narrative output
The monthly narrative is a generated draft, not a curated truth. You can edit it line-by-line, regenerate sections, or delete the whole thing before saving it as an Annal. The AI is a draft author. The cell is the editor.
Drift detection is a suggestion, not a diagnosis
The coaching report’s "drift" flag looks at engagement patterns — mission completion rate, journal sentiment shifts, frequency of cell interaction. A drift flag is a prompt to consider professional support if you feel it’s warranted. It isn’t a clinical assessment.
Safety moderation
We automatically screen user-submitted free text (for example, custom sitter activities) for child-safety using an AI classifier before it appears, through the same AI Gateway. Only the text being checked is sent — never your name, account, or other data — and we keep only a hashed fingerprint of the text and the safe/unsafe verdict, not the text itself. This moderation ledger (hashes plus verdicts) is retained for at most 90 days and is purged when you delete your account. Where moderation applies to Ember content you submit, it runs only under your explicit adult-feature consent.
Voice-note transcription
When you record a voice note and turn on transcription, the audio is sent to a speech-to-text provider (OpenAI Whisper or Google Cloud Speech-to-Text) via our AI Gateway solely to produce a transcript; it is not used for training. Off by default; child accounts require parental permission.
Frequently asked questions
Can the AI read my private journal entries?
No. Personal entries are visible only to you and the AI never reads them. It only reads entries you've explicitly marked family-narrative-eligible and opted into the monthly narrative; the default for every entry is personal.
Does the AI ever see my After Dark or Ember content?
Not for the family narrative or the coaching report — the data layer doesn’t expose Ember content to those features at all. AI touches Ember content in only two cases, both under the explicit adult-feature consent you give (GDPR Art. 9(2)(a)): an AI-powered Ember feature you opt into, and child-safety moderation of text you submit in Ember, which keeps only a hashed fingerprint and a safe/unsafe verdict, never the text itself. We don’t process Ember content with AI for any other purpose, and never as training data.
Is my data used to train the AI?
No. Our AI runs through a vendor-neutral AI Gateway (Anthropic, OpenAI, and Google), and none of them may use your data to train their models without an explicit opt-in. We have not opted in, and won't without telling you first.
Is the drift flag a medical or clinical assessment?
No. The drift flag looks at engagement patterns like completion rate and journal sentiment, and is just a prompt to consider professional support if you feel it's warranted. It isn't a clinical assessment.