Legal
Last updated July 30, 2026
LatentQ stores account identifiers, verified email status, practice progress, drafts, submissions, study history, and membership entitlements needed to provide and secure the service. Submitted code and interview answers are not sent to product analytics.
The five-question diagnostic stores derived competency signals and bounded session state, not the typed or spoken answer text. Unclaimed guest diagnostic sessions are removed after the configured guest-retention period. When a guest result is claimed by an account, its derived result remains part of that account history until account deletion or the applicable account-retention policy.
When a logged-in user explicitly requests an AI strong-answer comparison in the diagnostic, the selected question and answer are sent transiently to the configured AI provider for that request. The provider response and raw diagnostic answer are not persisted by LatentQ.
Payments are processed by the configured payment provider. LatentQ stores provider customer, checkout, transaction, and subscription references plus entitlement status; it does not store full card details.
Necessary transactional messages cover verification, password recovery, payment, receipts, waitlist confirmation, and account access. Optional founder updates and beta invitations use a separate preference that can be changed in account settings. Disabling optional product communication does not disable necessary transactional email.
When the private interview simulation uses adaptive interviewer turns or rubric evaluation, bounded portions of the CV, job description, typed or transcribed answers, recent transcript, and submitted evidence needed for that request are processed transiently through the LatentQ backend and the configured AI provider. LatentQ does not persist those request bodies, provider responses, or raw transcripts in its database or application logs. The provider may still process request data under its own service terms and data-handling policy.
When live voice is available, the browser connects directly to OpenAI Realtime over WebRTC for interviewer audio, microphone audio, and live transcription. The LatentQ backend authenticates the session and exchanges the bounded connection description without relaying the call audio. It also maintains a server-side provider control connection for fixed configuration, call lifecycle, usage and cost guardrails, and hangup recovery. Provider events on that control connection are processed transiently; transcript strings and audio-delta content are deliberately discarded and are not logged or persisted by LatentQ. LatentQ does not record or store call audio or video. OpenAI processes the live audio under its service terms and configured project data controls. The active transcript is held in the current browser tab. If you enable Recover after refresh, the interview draft and transcript are also kept in that tab’s session storage; the recovery draft deliberately excludes the CV and job description.
To keep an interview coherent without retaining raw candidate answers, LatentQ may store structured interview metadata such as server-resolved question and topic references and whether a response was captured. This structured metadata is retained for up to 30 days by default. Content-free cost, call-lifecycle, lease, usage, and operational telemetry may be retained for up to 730 days and is linked to internal session, user, project, and provider-call records. It can include provider, model, and service tier; one-way hashes of bounded browser call-attempt identifiers and short-lived teardown tombstones; content-free event identifiers; token and audio-duration counters; microUSD amounts and accounting status; and latency, fallback, retry, cancellation, violation, error, and completion metrics. Uncertain billing reservations may remain until they can be reconciled. Access to these records is restricted to authorized simulation administrators, and administrative viewing and export are audit logged.
Optional browser analytics starts only after you choose Allow analytics. It may include sanitized page paths, a pseudonymous first-party visitor identifier, browser-session timing, broad device category, referrer domain, campaign tags, and product actions. When a consenting session later signs in, its first-touch website and campaign metadata may be associated with that account. The consent prompt disappears completely after your choice. You may change the saved choice here on the Privacy page. Declining clears the browser analytics identifier and stops future optional browser collection.
Necessary server records continue to record account creation, verification, successful login, practice progress, submissions, promotional entitlement results, and membership state so the service can operate, remain secure, and provide account history. Promotional-code plaintext is never stored in analytics. Automatic token refreshes do not count as product activity.
Analytics events never include email addresses, passwords, tokens, promotional-code plaintext or hashes, raw IP addresses, full user-agent strings, source code, draft or notebook contents, custom cases, typed or spoken interview answers, raw runner output, card data, or payment instruments. Account identifiers already needed by the service may link authenticated activity to your account.
Detailed product events, pseudonymous sessions, first-touch attribution, promotional redemption attempts, and user-day activity are retained for 13 months. Minimal D1, D3, D7, and D30 return indicators remain with the account so its history stays accurate after detailed rows expire. Successful promotional grants remain necessary membership records. Non-identifying daily and retention totals may be retained longer. Read access and safe metadata exports are restricted to the single configured, verified analytics administrator and are audit logged; administrator audit records are retained for 24 months. Billing records follow separate account and payment retention requirements.
Configured infrastructure, email, payment, and analytics providers process only the information needed for their role. Access controls, verified webhooks, database constraints, and abuse protections are used to protect accounts and entitlements.
Privacy questions may be sent to contact@latentq.com.