Documents and scans
Forms, statements and claims with sensitive information alongside business content you need to retain.
PDFScans
GRITREDACT / CONTROLLED REAL DATA
Redact PII and PHI from documents, images and audio.
Detect and redact personal, health and business-sensitive information in documents, images and audio. Review what changes, preserve useful context, and export the result with an audit trail.
SENSITIVE SOURCE
↓ REVIEW & REDACT
CONTROLLED OUTPUT
The workflow
Detect. Review. Redact. Verify. Audit.
Bring in production files, identify sensitive entities, and apply the policy for the intended use. Human review stays part of the workflow before permanent removal.
On-premises: Docker in your environment Fully offline No network egress Local model inference
01Upload files in batches or connect through the API
02Detect PII, PHI and custom entities
03Review detections and apply your policy
04Redact or replace approved information
05Validate and export with an audit trail
What you can process
Images and identity records
ID cards, KYC images and photographed documents.
JPEGPNGTIFFBMPGIFWebP
Audio recordings
Redact sensitive speech in calls, voicemails and dictated notes while retaining the surrounding recording.
WAVMP3M4AOPUS
Detection, control and evidence
Local model
A fine-tuned open-source SLM runs inside your enterprise environment, without external model API calls.
Configurable entity policies
Choose which PII, PHI and custom entities to remove. Treat the same person or entity consistently across related files.
Human review
Review detections, correct selections and approve permanent removal. Operator decisions are recorded.
Removal and replacement
Mask or replace values according to your policy, including synthetic replacements. Pixel-level redaction permanently removes approved document and image content.
Preserved context
Keep document layouts, tables and surrounding content. Audio retains timing, speaker channels and surrounding sound.
Audit and export
Export sanitized files with a local record of detections, actions and locations, plus input/output hashes and operator history.
Where it fits
Use real content without sharing every detail.
Prepare approved material for development, QA, research, support escalation and AI or RAG ingestion. Keep the structure and real-world variation that simplified mock data misses.
For a larger test corpus, use reviewed output as a template in GritRender. When the next step requires structured fields, connect the document workflow with GritTransform.
- KYC documents, statements and customer call recordingsBanking
- Scanned forms, patient ID images and clinical audioHealthcare
- Claims, supporting images and first-notice-of-loss callsInsurance
- Purpose-specific copies for vendors and distributed teamsSharing
Choose how you deploy
Install Docker in your environment for fully offline processing with no network egress, or use our sales-assisted SaaS subscription through the web platform and APIs.
Enterprise deployment options include your data center, private cloud or VPC, Kubernetes and air-gapped environments. We’ll work with your team on the deployment and data flow.
Enterprise access and security
Access controls
Role-based access, SSO integration and tamper-evident records of operator actions and redaction decisions.
Integration
Review in the web platform. Connect through REST APIs, batch processing or watch folders for recurring intake.
Security review
Architecture and deployment reviews, security questionnaires, and support for penetration testing in your environment.
Before you get started
Does on-premises processing send files to a cloud model?
No. Model inference and processing run locally in the offline enterprise deployment. Documents, images and audio stay inside your environment.
Does redaction replace access controls?
No. If a user or agent is not entitled to a record, retrieval should be blocked. Redaction applies once the permitted purpose and minimum necessary content are known.
Will GritWorks train on our data?
GritWorks does not train models on client data unless that work is explicitly scoped and agreed to in writing.
How do we evaluate it for our workflow?
Bring representative file types, the entities you need to protect and the intended downstream use. We’ll walk through detection, review, output and deployment with your team.
Evaluate the workflow
Start with evidence you can inspect.
Use the evaluation guide to define the source content, review steps and output checks for your intended use.
Document redaction evaluation checklist
Review detection, permanent removal, retained context and audit evidence.
Loan-document testing with redact-then-generate
See the existing anonymized financial-services workflow.
Handwritten clinical content for AI testing
See how reviewed source content was prepared for controlled evaluation.
START WITH DOCUMENTS YOU CAN EXPLORE
Bring a sample into your workflow.
Free PDF packs for evals, sandboxes and testing: readable and degraded documents with reference fields. Need implementation or operational help? Discuss support around GritWorks products.
Start with your workflow
Put your unstructured data to work.
Tell us what your team needs to achieve. We’ll help you choose the right products, work through your requirements and explore on-premises or SaaS deployment.
Request a working session