01What AI features do
- Draft or remix content from a user's brief, brand/business context, and selected template text.
- Classify content and organize content ideas.
- Extract visible text and layout information from images through OCR.
- Where separately enabled and consented in the future, derive materials from meeting transcripts.
02What a model may receive
For a requested feature, input can include text briefs, template text, business/brand context, OCR text, or an image. The selected model may receive the material through OpenRouter or directly through a user-configured OpenAI, Anthropic, or Google BYOK connection.
Do not submit sensitive, confidential, or third-party personal information unless you have a valid reason and authority. The product cannot guarantee a model will identify or remove every personal-data element.
03Limits that remain unverified
- A product-wide redaction/minimization layer for all model inputs is not verified.
- Zero-retention and no-training commitments are not verified for every active model/provider route.
- Meeting transcript handling has identified historical plaintext storage and model-processing gaps; it is not approved for production Meeting Intelligence.
- Generated text, classifications, and OCR can be wrong, incomplete, biased, or unsuitable and must be reviewed before use or publication.
04Human control and Meeting Intelligence
Users choose whether to submit a brief, upload/select an image, connect a BYOK provider, or invoke an AI feature. Generated content should be reviewed and can be edited before use.
Meeting Intelligence is OFF without current, explicit, provider-scoped user consent. A separately initiated Spott/Fireflies BYOK creative import does not turn it on.
05Questions
Privacy/AI contact: privacy@viralz.dev (interim contact pending counsel approval of final privacy terms and operating procedures).