Enterprise Legal Data Safeguards

Security & Compliance at Disclosure Assistant

Institutional-grade protection engineered for legal practices handling privileged financial records, bank statement discovery, and court disclosures.

Last updated: September 17, 2026

Our Commitment to Legal Confidentiality

At Disclosure Assistant, protecting attorney-client work product and confidential financial records is our highest architectural mandate. We understand that bank statements, account ledgers, and cash flow records processed in our system represent sensitive discovery materials subject to strict legal privilege and ethical confidentiality rules.

Our platform operates on uncompromising principles of data minimization, zero artificial intelligence model training on customer files, and total cryptographic isolation between law firms. We do not sell, monetize, or repurpose user documents under any circumstances.

Every architectural layer—from file ingestion to database queries—is built to ensure that confidential client financial records remain accessible solely to authorized legal counsel.

Data Security Architecture

Encryption in Transit (TLS 1.3)

All data exchanged between client browsers, edge compute routers, and database instances is encrypted using Transport Layer Security (TLS 1.3 / TLS 1.2) with strict modern cipher suites. Insecure HTTP connections are automatically redirected to HTTPS via HSTS headers.

Encryption at Rest (AES-256)

Uploaded bank statement PDFs, parsed transaction records, and database rows are encrypted at rest using industry-standard AES-256 encryption. Encryption keys are managed and rotated via FIPS 140-2 compliant hardware security modules.

Row Level Security (RLS) on All Tables

Our PostgreSQL database enforces Row Level Security (RLS) directly at the database engine level across all tables (cases, transactions, case_files, integrations). Each database query is scoped strictly to the authenticated user ID (auth.uid()), mathematically preventing cross-tenant data leakage.

Continuous Backups & PITR

Database states are continuously backed up via write-ahead logging (WAL) with point-in-time recovery (PITR) capabilities. Backup archives are encrypted and geographically distributed across fault-isolated availability zones to guarantee disaster resilience.

Primary Data Hosting Region: Our core database, encrypted object storage, and serverless compute infrastructure are deployed in the United States (AWS us-east-1, Northern Virginia) through Supabase and Vercel. All client financial data resides within this region.

Infrastructure & Third-Party Subprocessors

Disclosure Assistant engages a limited set of verified, SOC 2 / ISO-compliant service providers to deliver hosting, database storage, email notifications, and payment processing.

Subprocessor / VendorFile Processing ScopePurpose in ArchitectureLocationData Accessed
Vercel, Inc.Processes File Stream (Memory)Application hosting, serverless compute runtime, global edge routingUnited States (US-East) / Global EdgeEphemeral HTTP payloads, client IP addresses; temporary file upload streams in memory
Supabase, Inc. (AWS)Stores Files & DB (AES-256)Managed PostgreSQL database, user authentication, and encrypted object storageUnited States (AWS us-east-1)User accounts, case names, parsed financial transactions, uploaded statement PDFs
OpenAI, LLCProcesses Text Lines (Zero Training)Table reconstruction & semantic transaction categorization (API)United StatesText-layer transaction lines (date, description, amount). Commercial zero-retention API; never used for model training
OkraPDFScanned PDF OCR FallbackProcesses Image Buffers (Memory)Optical character recognition (OCR) and layout extraction for scanned paper statements lacking a digital text layerUnited StatesDocument image buffers of scanned bank statements. Ephemeral processing; zero model training
Paddle.com Market LtdNo File Access (Billing)Primary Merchant of Record, subscription billing, and tax complianceUnited Kingdom / United StatesBilling contact details, subscription status, payment methods (PCI-DSS Level 1 compliant; raw card data never touches our servers)
CreemNo File Access (Billing)Secondary checkout and payment processing providerUnited StatesCustomer billing email, checkout session tokens, payment transaction confirmation
Resend, Inc.No File Access (Email)Transactional email deliveryUnited StatesUser email addresses, password reset tokens, security notifications
PostHog, Inc.No File Access (Telemetry)Product analytics and user session diagnostic telemetryUnited States (PostHog US Cloud)User IDs, email addresses, browser/device metadata, UI interaction events. Statement file contents are strictly excluded
Clio (Themis Solutions)Optional IntegrationUser-Initiated ExportDirect export of completed review reports into law firm matter foldersMulti-Region (US, CA, EU, AU based on firm setup)OAuth tokens, matter IDs, and exported report PDFs/CSVs when explicitly dispatched by user

Access Control & Engineering Hygiene

Mandatory Two-Factor Authentication (2FA)

Hardware-backed or authenticator-based 2FA is strictly enforced across all internal accounts, code repositories, cloud infrastructure consoles, and administrative services.

Principle of Least Privilege (PoLP)

Access to production systems is restricted strictly to programmatic API roles. Engineering personnel do not hold standing administrative database access; any operational maintenance requires documented, time-bounded elevation.

Protected Branches & Mandatory Code Reviews

All source code is maintained under strict GitHub branch protection rules. No direct commits to production branches are permitted; every pull request requires automated type checking, linting, and peer code review before deployment.

Strict Environment Isolation

Development, staging, and local environments operate against synthetic test fixtures and zero-value mock statements. Live production client data and uploaded statements are never cloned or mirrored to test environments.

Data Retention, Deletion & Subprocessor Disclosure

Exact specifications regarding stored data, retention schedules, instant deletion mechanics, disaster recovery backup policies, and subprocessor file handling.

1. What Data Is Stored

Disclosure Assistant operates on strict data minimization principles. We store only the discrete data points required to operate authenticated legal accounts, parse statements, and run disclosure checks:

Account & Auth Credentials

Attorney/user business email, encrypted password hash (managed via Supabase Auth; plaintext passwords are never received or stored), and authenticated session tokens.

Uploaded Statement Documents

Original bank statement PDF files uploaded by your firm, stored in private AES-256 encrypted storage buckets protected by PostgreSQL Row-Level Security.

Extracted Financial Ledger Rows

Normalized dates, extracted payees/descriptions, debits, credits, running balances, account numbers/suffixes, and exact source PDF coordinates (page number and line row).

Case Work-Product & Review Rules

Case matter names, custom dollar thresholds (cash limits, transfer limits), category flags, review statuses (Reviewed / Follow-up), and attorney internal notes.

Billing Records (No Raw Card Data)

Merchant of Record customer identifiers (Paddle/Creem) and subscription status. Credit card numbers, expiration dates, and CVVs are handled directly by Paddle/Creem under PCI-DSS Level 1 compliance and never touch our servers.

Integration Tokens (Optional)

Encrypted OAuth tokens for Clio Manage when connected by the user to push completed PDF/CSV exhibits directly into firm matter documents.

2. How Long Data Is Stored

Data retention is governed by customer account lifecycle and strict operational schedules:

  • Active Case Retention (Customer-Governed): Statement PDFs, parsed transactions, and attorney review notes remain accessible in your account for as long as you maintain the case active. You decide how long files remain in the platform.
  • Ephemeral In-Memory Processing: During extraction, statements processed through OCR (OkraPDF fallback) or table reconstruction APIs (OpenAI commercial endpoint) reside in volatile server memory strictly for the duration of the request (seconds) and are discarded immediately upon completion. No file or text cache persists on external compute.
  • Dormant Account Purge Policy: Accounts with cancelled subscriptions that remain inactive with zero user logins for twelve (12) consecutive months are scheduled for automated database and storage deletion, preceded by advance email notifications sent at 30 days and 7 days prior to deletion.
  • Session Inactivity Expiration: Browser sessions automatically terminate and lock after 15 minutes of inactivity to prevent unauthorized access on unattended law firm workstations.
3. How Users Delete Data (Self-Serve & Immediate)

Disclosure Assistant provides comprehensive, instantaneous self-serve deletion controls directly in the application interface:

Delete Individual Statement

Inside the Statements Drawer, click "Remove / Delete" on any uploaded file. This immediately deletes the PDF from encrypted storage buckets and cascades to permanently wipe all extracted transactions associated with that statement.

Delete Entire Case / Matter

Inside Case Settings, select "Delete Case". This immediately triggers a full cascading database purge that deletes all uploaded statement PDFs, all parsed transactions across every account, all notes, and the case record.

Delete Account & All Data

Inside Account Settings, confirm your password and click "Delete Account". This immediately revokes all authentication tokens, cancels subscriptions, and permanently purges all cases, files, ledgers, and profile records from production databases.

4. Whether Backups Retain Deleted Data (30-Day Expungement Cycle)

How disaster recovery backups handle deleted records:

Active Production Databases: When an item, case, or account is deleted by a user, it is immediately and irrevocably deleted from active production database tables and primary encrypted object storage buckets.

Disaster Recovery Snapshots: For business continuity and catastrophic disaster recovery, encrypted database snapshots and write-ahead logs (WAL) operate on a rolling 30-day lifecycle. Backup archives older than 30 days are automatically rotated and permanently overwritten/expunged.

Restoration Isolation Guarantee: Disaster recovery snapshots are strictly offline and isolated; they are never queried, accessed, or restored for normal operations or discovery. In the catastrophic event of a disaster recovery restore, active audit logs are reconciled immediately to ensure previously deleted customer data is promptly re-purged before production traffic is re-enabled.

5. Which Subprocessors Process Files

Only a strictly designated subset of subprocessors ever touches statement file contents or financial data:

Supabase / AWS:
Stores uploaded bank statement PDFs, parsed transactions, and case metadata at rest with AES-256 encryption in US-East (Northern Virginia).
Vercel, Inc.:
Hosts serverless application runtime. Streams file upload payloads in ephemeral server memory during extraction. No permanent file persistence.
OpenAI, LLC:
Receives raw text line items extracted from statements to assist in table reconstruction and layout parsing. Operates under commercial Zero-Data-Retention agreements; never used for model training.
OkraPDF:
Optical character recognition (OCR) fallback used strictly for scanned paper PDF files that lack digital text layers. Processes image buffers in volatile memory; zero permanent file storage; zero AI training.
Clio Manage (Optional):
Receives generated Completeness Audit PDFs and CSV files only when the user explicitly clicks "Upload to Clio".

Non-file subprocessors: Paddle and Creem (billing/payments only), Resend (transactional email only), and PostHog (UI interaction telemetry; statement contents and financial figures are strictly excluded).

6. Whether Files Are Used for Model Training (Contractual Zero-Training Guarantee)

Customer files, bank statements, transaction figures, notes, and metadata are NEVER used to train, retrain, fine-tune, or improve artificial intelligence models.

We contractually ensure that your uploaded bank statements, OCR text buffers, transaction amounts, payees, running balances, and attorney work-product notes are never used to train, fine-tune, or improve any public, commercial, or proprietary machine learning or AI models (including OpenAI foundational models). All commercial API integrations enforce zero data retention for training.

Incident Response & Breach Notification

Disclosure Assistant maintains a documented Incident Response (IR) Plan covering incident detection, threat isolation, forensic auditing, containment, and recovery.

24-Hour Notification Commitment: In the unlikely event of a verified security incident resulting in unauthorized access to, exposure of, or alteration of customer data, Disclosure Assistant will notify affected account owners via email within 24 hours of incident confirmation.

Breach communications provide actionable information including the scope of data involved, remediation steps implemented, and guidance for counsel.

Regulatory Compliance & Standards

GDPR Compliance

Full support for European and UK data protection standards, including individual rights to access, rectification, portability (CSV/JSON export), and erasure ("Right to be Forgotten").

CCPA / CPRA Compliance

Adherence to California consumer privacy requirements. We do not sell, rent, or monetize client personal or financial information.

PCI-DSS Level 1

All payment card handling is outsourced directly to certified PCI-DSS Level 1 Service Providers (Paddle & Creem). Our infrastructure never ingests or stores card numbers.

Security Questions & Vulnerability Reporting

Direct channel to our engineering and security response team

For security inquiries, vendor due diligence questionnaires, or to report a suspected vulnerability under responsible disclosure principles, please contact our security team:

Contact Security Teamsecurity@disclosureassistant.com