Privacy Policy (draft outline)
Updated October 10, 2026
Draft — pending legal review; not yet in effect
This document is a working draft prepared for counsel. It may change before launch and is not yet in effect.
Draft for legal review — not in effect.
This outline lists the personal data InfluenceBot actually stores, derived from the database schema (packages/database/prisma/schema/*.prisma) and the code, so counsel can draft the policy. Text in [COUNSEL: …] needs legal drafting. Items marked [Eng: …] are gaps the product must close before the policy can promise them.
1. Who we are
- Controller for account and billing data:
[COUNSEL: legal entity, address, contact, EU/UK representative if needed, DPO if needed]. - For content customers put into their workspaces (their audiences, leads, testimonials, real people in digital twins), InfluenceBot acts as processor; see the Data processing addendum.
[COUNSEL: confirm controller/processor split].
2. Personal data we collect
2.1 Account and sign-in (controller)
- User: email, name, optional profile image URL, language, time zone, status, last sign-in time.
- Sessions: a hashed session token, expiry, IP address and user agent, the active workspace; platform-staff impersonation fields (who impersonated and until when).
- Sign-in accounts: password hash; for OAuth sign-in, provider account id and encrypted access, refresh and id tokens.
- Verification tokens for emailed links (stored hashed).
- Tables for two-factor secrets (encrypted) and passkeys exist; no user interface uses them today.
- Workspace membership, roles, invitations (invitee email, inviter, expiry), notifications.
- Enterprise SSO and SCIM: IdP subject, user name, email, display name, group membership, SSO event log with IP address.
- API keys: prefix, key hash, scopes, last-used IP address.
2.2 Billing (controller)
- Billing customer: Stripe customer id, billing email, a reference to the default payment method (card data stays with Stripe), tax details.
- Subscriptions, seats, credit ledger, credit-pack purchases, coupons and referrals, enterprise invoices.
- Trial-abuse signals: keyed hashes of IP address and device identifier (not the raw values).
2.3 Audit and security (controller)
- Audit log of actions: actor, action, resource, IP address, user agent, correlation id, metadata. Append-only.
- Rate-limit counters keyed by user, session or IP (Redis, short-lived).
2.4 Content in workspaces (processor for the customer)
- Brand knowledge: website pages and documents the customer imports (may contain personal data of the customer's staff).
- Personas: for a digital twin of a real person — the subject's legal name, relationship, consent type and status, allowed uses, territories, release document, expiry and revocation; reference images; voice settings (voice cloning is disabled until consent is recorded).
- Customer testimonials: customer name, quote, source URL, linked consent.
- Generated and uploaded assets (images, audio, video, documents) with provenance.
- Approvals, comments, QA reports.
- Copilot conversations and messages; MCP tool-call logs (inputs/outputs or their hashes); automation activity.
2.5 Social accounts and audiences (processor)
- Social connections: network account id, name, handle, encrypted OAuth tokens, scopes, connected-by user.
- Published posts and metrics; inbox threads and messages (author handle, message body) where a network supports them.
- Listening mentions: author handle (and a hash), post text, sentiment. Today only Mock listening connectors run; real connectors are Coming soon.
2.6 Leads, links and commerce (processor)
- Leads from public lead forms: email, name, phone, consent records per purpose (append-only, with a hash of the consent text and keyed hashes of IP and user agent), double opt-in, withdrawal.
- Tracking-link clicks: a keyed hash of IP + user agent with a rotating salt, the user agent, referrer, device type, bot flag. No cookie is set on the visitor. The destination URL gets an
ib_clickparameter. - Affiliate-link clicks: visitor hash, user agent, referrer host, bot flag.
- Conversions and imported commerce orders (amounts; external ids).
- Sponsors, deals, usage rights, media kits.
2.7 Support and services (controller/processor)
- Support tickets (subject, body), service orders, intake form submissions, messages with staff, staff access grants.
3. Purposes and legal bases
[COUNSEL: map each category above to a purpose and legal basis — contract, legitimate interests (security, fraud/trial abuse, audit), consent (marketing, cookies if any), legal obligation (tax, invoices)].
4. AI processing
- Prompts and workspace content (brand knowledge, scripts, product images, persona references) are sent to the AI provider chosen for each task; outputs are copied into the workspace storage. See Sub-processors for each provider and its status.
- The platform blocks Live use of a provider until its commercial and data terms are reviewed and recorded.
[COUNSEL: state whether customer content is used to train models — Eng: the product does not train models on customer content; provider-side retention depends on each provider's terms].- Automated recommendations (portfolio decisions) are never applied without a human accepting them.
[COUNSEL: automated decision-making statement].
5. Sharing
- Sub-processors (hosting, AI providers, payments): see Sub-processors.
- Social networks and commerce tools the customer connects (customer-directed).
- Platform staff with time-limited, audited access grants for professional services or support.
[COUNSEL: legal requests, business transfers].
6. International transfers
- Hosting is on Google Cloud.
[COUNSEL: regions — Eng/Founder to confirm the Cloud SQL and bucket regions; transfer mechanisms for AI providers in the US].
7. Retention (behaviour in the code today)
| Data | Retention in code |
|---|---|
| Sessions | Expire after 7 days (refreshed at most daily) |
| Emailed sign-in links | 10 minutes, single use |
| Password reset links | 1 hour |
| Invitations | 7 days by default (up to 30) |
| Deleted assets | Soft-deleted, then purged after 30 days by default (configurable); assets still referenced by lineage or approvals are kept |
| Public API event log (webhooks) | 30 days |
| Analytics ingestion journal | 7 days by default |
| Background job records | Completed jobs about 1 hour, failed 7 days, dead-letter 30 days |
| Lead consent records | Append-only; kept with the lead |
| Audit log, credit ledger | Append-only; no deletion schedule in code [COUNSEL: retention period] |
| Everything else | No automatic deletion in code [COUNSEL + Eng: retention schedule] |
8. Your rights
[COUNSEL: access, rectification, erasure, restriction, portability, objection, withdrawal of consent, complaint to a supervisory authority].- Self-service rights (ADR-147,
docs/specs/privacy.md): any user can export their account data and request account deletion; workspace owners can export or delete a workspace. Exports are ZIP bundles kept 7 days behind a 15-minute signed link (3 per day). Deletions are confirmed, then wait a 14-day cancellable grace period; personal data is then erased or anonymised, while the credit ledger, billing records and audit log are kept and point to an anonymous id.[COUNSEL: retention periods for invoices, ledger, audit log and consent proofs; whether to describe the grace period]. - Lead consent withdrawal is supported (append-only record; list syncs stop at once).
- Persona consent can be revoked; affected assets are marked revoked and approvals are invalidated.
9. Security
- TLS in transit; OAuth tokens, provider keys, two-factor secrets and SSO secrets encrypted at rest; passwords hashed; API keys shown once and stored hashed; tenant isolation enforced in the data-access layer; audit logging.
[COUNSEL: wording].
10. Cookies
- See the Cookie policy.
11. Children
- The service is for business users. Persona workflows block minors' likeness.
[COUNSEL: minimum age statement].
12. Changes and contact
[COUNSEL: change notice, contact address — no privacy contact or support address is wired into the app yet].
