Overview#
Before a prosecutor hands over evidence to the defence, every piece of body camera footage and every interview transcript must be reviewed for faces, personal identifiers, and Brady material. Doing that manually across hundreds of files takes weeks and introduces human error. The Privacy domain automates the detection and redaction work, flags exculpatory evidence against Brady v. Maryland obligations, and generates a compliance report showing exactly what was reviewed and when.
Key Features#
- PII detection: SSN, phone, email, address, date of birth, credit card, licence, passport, and custom patterns
- ML-based recognition with confidence scoring and bounding box location
- Face detection with bounding boxes, classification (subject, witness, law enforcement), and demographic estimation
- Automated redaction: black box, blur, pixelate, text replacement, and removal methods
- Brady material identification: exculpatory, impeachment, mitigating, credibility, and misconduct evidence
- AI-based significance scoring with deadline tracking and disclosure audit trails
- Compliance with Brady v. Maryland disclosure obligations and CJIS security requirements
- GDPR and CCPA support for data minimisation and subject access requests
- Configurable confidence thresholds for detection and auto-redaction
- Comprehensive compliance scoring and review status tracking
Use Cases#
Relevant sectors include law enforcement, financial crime investigation, and intelligence agencies.
- Detecting and redacting PII in investigation documents before disclosure or sharing
- Automatically identifying and blurring non-subject faces in body camera and surveillance footage
- Scanning case evidence for Brady material to ensure timely prosecution disclosure obligations
- Generating compliance reports showing redaction and disclosure completion status
Integration#
The Privacy domain connects with evidence management, investigation context, export systems, and secure sharing workflows. It uses OCR engines for document processing and ML models for face detection and Brady analysis.
Open Standards#
- GDPR (EU Regulation 2016/679): Every privacy audit log entry carries explicit GDPR-compliance flags, and the domain supports data minimisation and subject access request workflows required by the regulation.
- EU AI Act (Regulation 2024/1689), Article 5(1)(g): Face-detection models are constrained by this provision, which prohibits inferring race or ethnicity from biometric data; the
ethnicityfield is deprecated and must not be populated. - CJIS Security Policy (FBI): Evidence redaction and audit-trail capabilities are designed to meet the Criminal Justice Information Services security requirements cited as a compliance target for law enforcement deployments.
- HIPAA (45 CFR Parts 160 and 164): PII classifications include medical record identifiers, and the
protected_classattribute on detections records HIPAA-protected categories to gate redaction and disclosure decisions. - CCPA (California Civil Code §1798.100 et seq.): Audit log records carry an explicit CCPA-compliance flag, and data-minimisation redaction templates support consumers' right-to-deletion obligations.
- GraphQL (June 2018 Specification): All PII detection, face detection, Brady material, and redaction operations are exposed as typed GraphQL queries and mutations via the Strawberry schema layer.
- SHA-256 (FIPS 180-4): Redaction placeholders for text content are generated using SHA-256 hashes rather than MD5, in line with NIST-approved hashing requirements for law enforcement evidence handling.
Last Reviewed: 2026-02-05 Last Updated: 2026-04-14