Data residency & retention
Your documents stay in the UK. We run API calls from a Microsoft data center in the EU.
Each workspace chooses how long converted outputs are kept: seven days by default, up to 30 days. They are then deleted automatically.
Security and privacy
Client documents, treated like client documents.
LexForge handles legal documents. This page sets out where your files are stored, what processes them, and what your firm controls.
UK document storageEU data processingNever used to train AI
Document controls
How LexForge handles your documents.
Wording and training
LexForge does not use customer documents to train any AI model.
Automatic scan cleanup
Your source scan goes on the short-retention path the moment it arrives, then leaves active storage automatically.
Retention your firm controls
Your workspace sets how long finished files stay available, from one to 30 days. Source scans always follow the shorter path.
Billing, email and support details sit with the sub-processors listed below.
Technical detail: what runs where, and for how long
Documents and OCR stay in Microsoft Azure UK South. For structure and layout analysis, the Azure OpenAI EU Data Zone within the EEA receives relevant full-page images and extracted text. Those requests run with storage disabled (store=false), so normal Responses API conversation state is not retained, and customer material is not used to train generative models. Microsoft applies automated abuse monitoring: material flagged by the abuse-monitoring system, or identified as part of a potentially abusive pattern of use, may be stored by Microsoft and reviewed by authorised Microsoft staff.
Uploaded scans are scheduled for deletion from active storage two hours after upload, If that cleanup is delayed, an Azure Storage lifecycle rule makes the source and extracted-text cache eligible for deletion once they are more than one day old; Microsoft controls the exact processing time. Deleted blobs may remain recoverable to authorised administrators for up to seven days under Azure soft delete.
During house-style setup, LexForge reads the supplied reference document to learn its formatting and structure. Reference material is handled under the same processing and retention safeguards.
AI can reconstruct structure and wording from page images. Wording must be corroborated by an independent reading of the source or confirmed by a person against the page image. Unresolved wording is held for review.
Sources: Microsoft on Azure OpenAI data, privacy and security, Azure OpenAI deployment types, Azure Storage lifecycle management and Azure blob soft delete. The full policy is in our privacy policy.
Where each kind of data lives, and for how long
| Data | Where it is processed | How long it is kept |
|---|---|---|
| Source scan (uploaded PDF or Word file) | UK | Scheduled to leave active storage two hours after upload. An Azure lifecycle rule backstops anything older than one day. |
| OCR reading (extracted text) | UK | The standard extracted-text cache is scheduled for deletion two hours after upload, with the same lifecycle backstop. |
| Page images and text sent for structure and layout analysis | EU | Nothing is retained beyond Microsoft's abuse monitoring described above. |
| Converted Word output, reconstruction report and saved comparison evidence | UK | Seven days by default, up to 30 days, set per workspace, then deleted automatically. |
| Error monitoring and product analytics | EU | Operational metadata only. No document text, filenames, email addresses, job IDs or tenant IDs. |
Illustrative change report
What the wording check proves.
Every reconstruction records the wording check and flags uncertain words. The example below is invented, but the layout is what your team sees.
Change reportExample document · 14 pages
Ready for review- Pages checked
- 14 / 14 ✓
- Words compared
- 9,412 ✓
- Words flagged for review
- 3
- Wording check
- Matched ✓
Data flow
The path of a document
Upload
Your file arrives in LexForge.
UKRead
Azure Document Intelligence reads the words.
UKUnderstand structure
AI reconstructs headings, numbering, layout and wording. Wording is checked against an independent source reading.
EUReturn and retain
You get an editable Word file. Your workspace decides how long it stays.
UKDue diligence
Sub-processors
These are the companies that process data for LexForge. Documents and the reading of them stay in the UK. Structure and layout analysis runs in the EU.
| Processor | Purpose | Region |
|---|---|---|
| Microsoft Azure | Hosting, storage, database, Document Intelligence OCR, document-processing services and transactional email delivery | UK |
| Azure OpenAI Service | Structure, layout and wording reconstruction from page images and extracted text, checked against an independent source reading. Storage-disabled requests; Microsoft abuse monitoring applies, as set out above. | EU |
| Stripe | Payment processing and billing | EU/UK |
| Clerk | Sign-in and user accounts (names, work emails) | EU |
| Sentry | Browser error monitoring for product pages; no document text, filenames, email addresses, job IDs or tenant IDs | EU |
| PostHog | Cookieless product analytics for coarse page and feature usage; no document text, filenames, email addresses, job IDs or tenant IDs | EU |
Security questionnaires
Need help with a data protection impact assessment (DPIA) or security questionnaire?
Send yours over, or ask for our security pack, and we will reply with the documents you need. Our data-flow diagram for DPIAs is available on request.
Prefer email? support@lexforge.app
Typical response within two working days.