Legal
AI disclosure
Last updated: October 2026
AutoArchive uses automated (AI) extraction to read information from documents you upload. This page explains what that does, how confident it is, and what stays under your control.
What the AI does
When you upload a photo or file — an MOT certificate, invoice, service record or similar — AutoArchive runs automated extraction to read values such as dates, mileage, amounts, part names and document type. This is designed to save you typing, not to make a decision for you.
Confidence and review
Every extracted value is shown to you with a confidence indicator before it is saved against your vehicle. Low-confidence values are flagged for your attention. Nothing extracted is treated as final until you have reviewed it.
Your corrections
If an extracted value is wrong, you correct it on the review screen. Your correction is saved alongside the original extraction and original document — the source file is never overwritten, so you can always see what was scanned.
What the AI does not do
- It does not diagnose mechanical faults or replace inspection by a qualified mechanic.
- It does not verify a document is genuine, or check figures against DVSA, DVLA or any external register beyond the MOT history lookup described below.
- It does not change or overwrite DVSA MOT history — MOT data comes directly from the DVSA MOT History API and is never AI-generated.
- It does not decide account, billing or workspace-role outcomes.
Duplicate and rescan detection
AutoArchive also uses automated similarity checks to flag when a newly uploaded document looks like an existing one (an exact duplicate or a clearer rescan). You choose whether to keep both, replace, or discard — the system never deletes a document automatically.
Where processing happens
Extraction runs in AutoArchive's server-side pipeline through the Lovable AI Gateway. The currently configured model is OpenAI GPT-6 Astra. Our request to the model asks for the response not to be stored. The document is processed first for structured extraction and then for verification; the result is returned to your account for review.
AI provider, training, retention and location
Provider: our platform and AI gateway provider is Lovable (Lovable Labs Incorporated, Boston, USA, with Lovable Labs AB, Stockholm, Sweden), acting as our processor under its Data Processing Agreement.
Training: section 10 of that agreement states that “Lovable shall not use any Customer Personal Data for the purpose of training, retraining, fine-tuning, or otherwise developing any Artificial Intelligence (AI) or Machine Learning (ML) models.”
Retention: the agreement says Lovable retains Customer Personal Data “as in Annex 1” and deletes or returns it when the agreement ends. Annex 1 does not state a specific retention period for AI requests.
Location: the agreement does not name the countries where AI processing takes place. Its sub-processors are listed at trust.lovable.dev.
Publication blocker
Draft — pending owner sign-off. Document scanning stays off on the published site until this page is approved.
AI Assistant
When you ask the AI Assistant a question, it sends your question and a summary of vehicles in the active workspace through the Lovable AI Gateway to Google Gemini 2.5 Flash. Answers may be incomplete or mistaken; check them against your records.
Connected third-party AI apps
The MCP/OAuth connector can let a third-party AI app read vehicle history when you authorise it. This is separate from document extraction and the AI Assistant. You can revoke the connection; the app's own data practices apply to what it has received. See our privacy policy.
Contact
Questions about automated extraction: contact@autoarchive.uk. See also Privacy policy.
