1. Active AI/OCR Configuration
Completa's launch AI/OCR configuration is a dual-provider model:
- Anthropic for project image understanding, classification support, scope analysis, estimate support, quality review, dispute/warranty support, and related project image analysis.
- OpenAI for receipt, invoice, materials-record, and document OCR/extraction where configured.
Privacy Policy, AI/OCR Disclosure, upload disclosures, and subprocessor records must identify both provider categories where active.
2. Approved Disclosure Language
Completa may use approved external AI/OCR processors, including Anthropic for project image understanding and OpenAI for receipt, invoice, material-record, and document OCR where configured. Completa sends original inputs only for the specific processing request, receives returned readings, keeps returned readings internal, and does not permit external providers to train general models on customer or contractor content.
Completa may use customer and contractor project data to improve Completa's own internal pricing engine, platform features, quality controls, dispute/warranty workflows, fraud/safety review, and support. This internal work is not external AI provider training. Completa does not send returned readings, extracted fields, summaries, classifications, or internal analyses back to an external provider for training, evaluation, benchmarking, feedback, or model improvement unless a later material policy update is accepted where required.
3. Provider Control Requirements
Each active provider must be covered by no-training or equivalent commercial/API terms, limited-retention terms where available, processor/vendor review, access controls, and internal logging of provider, purpose, file ID, project ID, request timestamp, and returned-reading storage location.
4. Prohibited Uses Without Reacceptance
- provider model training;
- sending returned readings back for evaluation/training/benchmarking;
- exposing raw AI readings as authoritative user-facing decisions;
- biometric identification, face recognition, voiceprint recognition, identity recognition, or hand/face geometry extraction;
- new provider with materially broader purpose;
- identifiable internal AI training beyond disclosed pricing-engine/platform improvement.
5. Control Commitments and SOP Alignment
Completa processes project data only through approved commercial/API provider configurations, not consumer AI accounts. Completa does not opt in to provider training, does not submit feedback containing customer or contractor data, applies access controls to original inputs and returned readings, and keeps provider logs.
Completa applies automated or procedural filtering before provider submission to redact payment-card and bank-account details from receipts/invoices, avoid irrelevant faces and children's images, and strip EXIF/GPS metadata from photos, except where the information is required for the specific processing request, legal retention, fraud/safety review, dispute/warranty review, or the user has been notified. Provider logs must identify the non-biometric project purpose for each image-processing request. Detailed implementation controls live in the AI/OCR processing SOP and must stay aligned with this addendum.