PreFlight AI1. Local-First Client Architecture vs. Ephemeral Cloud Verification
To protect pre-publication academic research, PreFlight AI isolates manuscript processing into two strictly defined execution zones:
Zone 1: Local-First Client Workspace
LaTeX parsing, mathematical-region isolation, bibliography extraction, and editing run in your browser. Local workspace recovery may use IndexedDB. Full draft files are not stored in our cloud databases; closing a tab does not necessarily clear local files.
Zone 2: Ephemeral Stateless Cloud Calls
Cloud verification can transmit isolated claim sentences and source identifiers over HTTPS. This is separate from full-draft parsing. User-linked cloud records may be retained under the configured verification path; independent providers have their own processing and retention policies.
2. AI Inference Architecture & Zero Training Commitments
Cloud-assisted checks are subject to the active provider and configuration:
- •Zero Model Training: PreFlight AI does not use manuscript data to train models. A no-training commitment is not a blanket promise of zero retention by every provider. Consult the relevant provider terms before processing sensitive claims.
- •Bring-Your-Own-Key (BYOK) Governance: Where a bring-your-own-key integration is supported, provider processing follows the terms for your own account. Do not assume that every inference path supports direct routing or zero retention.
- •Draft and verification boundaries: Full manuscript drafts must not enter application logs or cloud draft storage. Verification data is a separate data class and should not be treated as automatically erased at response completion.
3. Information We Collect & Store
We limit backend data collection strictly to what is necessary for authentication, billing, and system stability:
- •Account & Authentication Records: When you register or sign in via OAuth (Google, GitHub, or Microsoft), we store your email address, display name, user ID, and session tokens in Cloudflare D1 to manage access and license entitlements.
- •Subscription & Transaction Ledger: We maintain records of subscription status, cryptographic license tokens, seat allocations, and order identifiers. Payments are handled by Dodo Payments. PreFlight AI does not store card numbers or bank account credentials. Payment providers and legally required financial records have independent retention obligations.
- •Fair Usage & Free Trial Limits: Quota and abuse-prevention records may include counters and hashed network identifiers. Billing and metering records are distinct from manuscript drafts.
- •Operational Diagnostic Telemetry: Operational metadata such as request status and latency may be used to diagnose failures. Full draft content must not be included in diagnostic telemetry.
4. Scholarly Registries & Network Disclaimers
PreFlight AI cross-references bibliographic metadata against independent third-party public academic databases, including CrossRef, OpenAlex, Semantic Scholar, PubMed, and Retraction Watch.
- •Upstream academic registries are operated independently and may experience indexing delays, downtime, or incomplete metadata.
- •Transmission across public telecommunications networks is subject to standard Internet routing and encryption protocols (TLS 1.3).
- •PreFlight AI disclaims liability for the availability, completeness, or privacy practices of upstream external academic registries.
5. Prohibited Materials & Compliance Warranties
You agree not to upload, process, or verify with PreFlight AI any materials that constitute: (a) national defense secrets or export-controlled munitions technical data (ITAR/EAR); (b) Protected Health Information (PHI) under HIPAA; (c) unredacted personal financial data; or (d) non-consensual biometric or identifying data. PreFlight AI expressly disclaims liability for unauthorized submission of legally restricted materials.
6. Global Privacy Rights & Self-Serve Data Erasure (GDPR / CCPA)
Regardless of your geographic location, you maintain full control over your personal information:
- •Self-Serve Account Deletion: Account settings provides a confirmed account-deletion workflow. Covered application records are removed when deletion succeeds; incomplete or failed operations must not be treated as success. Independent providers and legally required payment records are outside this application deletion.
- •Local Workspace Reset: Use local-data controls in Account Settings or your browser site-data settings to remove workspace files. Browser settings can also remove preferences and cookies. Local clearing does not purge cloud records.
- •Cloud Data Purge: The authenticated
POST /api/user/purge-cloud-dataendpoint removes covered user-linked cloud records; it does not erase independent provider records, statutory billing records, or public paper caches. This endpoint is separate from local clearing and is not currently a settings button.
Privacy Inquiries: Questions regarding data governance or requests for personal data export may be directed to support@preflightai.tech.
7. Advisory Scope & Legal Non-Liability
By using PreFlight AI, you understand that PreFlight AI is an automated pre-submission decision-support tool. Authors maintain sole, final responsibility for the veracity, ethical compliance, and submission of their manuscripts. PreFlight AI assumes no liability for editorial outcomes, journal peer-review decisions, or third-party telecommunication anomalies.
Your continued use of PreFlight AI indicates your informed acceptance of this Privacy Policy and Data Governance agreement.