Security and trust

Built for confidential disability-law records.

DisabilityProAI processes medical records, legal work product, OCR text, embeddings, reports, and case chat. Our security program is designed around strict access control, short-lived document access, vendor-backed cloud safeguards, independent Thoropass penetration testing, and controlled production change management.

HIPAA-conscious legal workflows

Designed to support sensitive U.S. disability-law workflows when appropriate agreements and customer-side controls are in place.

Thoropass pentest completed

Thoropass independently tested the web app and API, and its remediation review confirmed the reported items were resolved.

Security white paper

A customer-facing white paper summarizes architecture, data handling, subprocessors, delivery controls, and the Thoropass attestation.

No model training

Customer content is used to provide the service, not to train foundation models or customer-specific AI models.

Security white paper

Download the law-firm security overview covering data handling, infrastructure, access controls, subprocessors, retention, incident response, and Thoropass penetration testing.

DisabilityProAI-Security-White-Paper-2026-06-04.pdf

Download PDF

Thoropass pentest attestation

Download the Thoropass letter attesting to independent web app and API penetration testing, remediation, and remediation-review completion.

DisabilityProAI-Thoropass-Penetration-Test-Attestation-2026-06-02.pdf

Download PDF

Data protection

How case data is protected

Uploaded documents and generated work product are treated as sensitive customer content. Access is intentionally narrow, logged where practical, and limited to what is needed to provide the service.

No model training on customer content

DisabilityProAI does not train foundation models or customer-specific AI models on uploaded files, OCR text, prompts, embeddings, reports, or chat content.

Encrypted transport

Application traffic is served over HTTPS with HSTS, and backend service calls use authenticated TLS connections.

Encrypted storage

Primary data stores and object storage rely on managed cloud encryption at rest, with application controls layered above provider safeguards.

Short-lived access

Reports and source-file views use authorization checks and signed access patterns instead of public file locations. Browser uploads use server-issued, time-bound Google Cloud upload sessions.

Application controls

Access is enforced server-side

The application does not rely on hidden IDs, browser-only checks, or UI visibility as security boundaries.

Supabase Auth plus ownership checks

Users authenticate through Supabase Auth. Sensitive API routes verify the authenticated user and the specific case, file, or billing resource being requested.

Row-level and role-based boundaries

Customer-facing roles are scoped so users can access their own matters, while prompt and admin configuration is handled through protected server routes.

Admin session hardening

Admin-only tools require explicit authorization, stale admin sessions are denied, and admin activity is recorded for review.

Bot and abuse controls

Authentication-sensitive endpoints fail closed without server-side hCaptcha verification, with rate limiting on auth paths and high-cost AI/document-processing endpoints.

Browser attack reduction

The app sets CSP, HSTS, X-Frame-Options, Referrer-Policy, and Permissions-Policy headers, with Origin checks on state-changing API requests.

AI prompt protection

Proprietary prompts and workflow instructions are kept out of public bundles and ordinary browser-readable Supabase responses.

Third-party testing

Thoropass penetration testing completed

Thoropass performed independent gray-box penetration testing of the DisabilityProAI web application and API from April 28, 2026 to April 30, 2026. The public attestation states that all findings identified during the assessment were remediated and that a remediation review confirmed the reported items were resolved.

Independent gray-box assessment

Thoropass assessed the DisabilityProAI web application and API from April 28, 2026 to April 30, 2026 using pre-provided credentials and limited knowledge.

Industry-standard methodology

The attestation references OWASP Top 10, OWASP API Security Top 10, applicable OWASP WSTG sections, and relevant NIST SP 800-115 assessment guidance.

Web and API attack coverage

Testing covered authorization, authentication, session and JWT handling, injection, file handling, rate limiting, sensitive data exposure, security configuration, and business-logic abuse classes.

Remediation review passed

Thoropass reported that all findings identified during the assessment were remediated and that a remediation review confirmed the reported items were resolved.

Infrastructure

Vercel, Supabase, and Google Cloud

DisabilityProAI uses managed cloud infrastructure with a clear split between the web app, data layer, long-running document-processing worker, and payment/support systems. Secret keys stay server-side, and backend services use authenticated service-to-service calls.

1

Vercel application layer

The Next.js application is deployed through Vercel with immutable builds, production branch protection, managed TLS, edge routing, rollback support, bot protection, firewall rules, and rate-limit capabilities.

2

Supabase data layer

Supabase provides Postgres, Auth, private Storage, and pgvector retrieval. Server routes use the service role only on the backend; customer-facing access is constrained by auth, ownership checks, grants, and RLS.

3

Google Cloud processing layer

Long-running OCR, embedding, PDF generation, extraction, organization, and formatting work runs through Google Cloud Run, Cloud Tasks, Cloud Storage, and Vertex AI with service-to-service authentication.

4

Separated staging lane

A persistent Supabase staging branch and staging Cloud Run worker are used for structural tests before production database or worker changes.

Operations

Secure delivery and change control

Production changes go through a protected software delivery flow, with automated checks, production bundle scans, and an audit trail for security-relevant changes.

Private source repositories and protected production branch.
CodeQL, Dependabot, Socket Security, and Vercel checks on production changes.
GitHub secret scanning with push protection for credential exposure prevention.
Release checks for dependency risk, npm install-script review, production bundle secret scanning, and npm audit.
Change review, rollback support, and controlled production deployment procedures.
Customer assurance materials maintained for review under appropriate confidentiality terms.

Customer documentation

DPA and customer assurance materials

Law firms often need more than a security page. We are organizing a standard documentation package that can be reviewed by firm leadership, IT, compliance, or outside counsel. The preferred onboarding flow is click-through acceptance with the current Data Processing Agreement visible below and a signed PDF copy emailed after subscription checkout; custom signatures or a BAA can be handled on request.

Data Processing Agreement

Our standard Data Processing Agreement covers processing purposes, data categories, subprocessors, deletion, incident notice, audit support, customer instructions, and the no-sale and no-model-training commitments.

Subprocessor list

Core subprocessors include Vercel, Supabase, Google Cloud / Vertex AI, Stripe, and communication providers used for service emails and support.

HIPAA-conscious workflows

The service is designed to support sensitive disability-law workflows when required agreements and customer-side controls are in place.

BAA on request

A Business Associate Addendum is separate from the DPA and should be used only when a customer needs HIPAA business associate terms for an approved workflow.

Standard DPA

Data Processing Agreement

The full standard DPA is shown here for review. A PDF copy signed electronically by Canary Doctor LLC dba DisabilityProAI is available for download and is also emailed after subscription checkout.

Signed PDF copy

DisabilityProAI-Data-Processing-Agreement-2026-05-22.pdf

Download PDF

DisabilityProAI Data Processing Agreement

Last updated: 2026-05-22

1. Parties And Scope

This Data Processing Agreement ("DPA") forms part of the agreement between Canary Doctor LLC dba DisabilityProAI ("Provider") and the customer identified in the applicable order form, subscription, or terms of service ("Customer").

This DPA applies when Provider processes Customer Content on behalf of Customer through the DisabilityProAI service. If there is a conflict between this DPA and other agreement terms about processing Customer Content, this DPA controls for that subject matter unless the parties expressly agree otherwise in writing.

The service and this DPA are designed for United States customers, United States-based Authorized Users, and United States disability-law workflows only.

2. Definitions

  • Authorized User means a person Customer permits to access the service.
  • Customer Content means files, text, case metadata, OCR output, embeddings, prompts, reports, chat content, instructions, and other information submitted to or generated through the service for Customer.
  • Personal Data means Customer Content or account/support information that identifies or can reasonably be linked to an individual.
  • Sensitive Data means Personal Data that includes medical records, health information, disability records, government-benefit materials, attorney work product, financial information, or similar sensitive information.
  • Security Incident means a confirmed unauthorized access to or disclosure, alteration, or destruction of Customer Content in Provider systems.
  • Subprocessor means a third party engaged by Provider to process Customer Content or Personal Data to provide, secure, support, or maintain the service.

3. Roles And Instructions

Customer determines what Customer Content is submitted to the service and is responsible for having the rights, permissions, notices, and authorizations needed to use the service.

Provider will process Customer Content only to provide, secure, maintain, support, and troubleshoot the service as permitted by the agreement, Customer's product configuration, Customer's support requests, and lawful written instructions from Customer.

Provider will not train foundation models or customer-specific AI models on Customer Content.

Provider does not sell Customer Content or Personal Data. Provider does not share Customer Content for cross-context behavioral advertising, marketing, or unrelated third-party use. Disclosures to subprocessors listed in this DPA are for service delivery, security, support, billing, or legal compliance, not for sale of Customer Content.

Provider may decline or pause an instruction if Provider reasonably believes it is unlawful, technically infeasible, would compromise security, or would create material risk to the service or another customer.

4. Confidentiality

Provider will restrict personnel access to Customer Content to personnel and contractors with a business need to know. Provider personnel and contractors with access to Customer Content will be subject to confidentiality obligations.

Support access to Customer Content should be limited to the access needed to resolve a support, security, billing, or operational issue. Customer should not send PHI, medical records, or other Sensitive Data through email or support channels unless Provider has explicitly approved that channel for such data.

5. Security Measures

Provider will maintain reasonable technical and organizational safeguards designed to protect Customer Content. Current safeguards include the measures in Appendix 1. Provider may update safeguards over time, provided updates do not materially reduce the overall protection of Customer Content.

6. Subprocessors

Customer authorizes Provider to use the following subprocessors to provide, secure, support, maintain, and bill for the service:

SubprocessorPurposeData categoriesNotes
VercelNext.js application hosting, deployment, edge routing, TLS termination, firewall/rate-limit controls, and deployment logs.Account/session metadata, request metadata, application logs; Customer Content may transit serverless functions as part of service delivery.Used for the public web application and protected application routes.
SupabasePostgres database, authentication, private storage, pgvector retrieval, and operational logs.Account data, authentication data, Customer Content, case metadata, OCR text, embeddings, generated outputs, usage logs.Used for managed database, authentication, storage, and retrieval services.
GoogleGoogle Cloud Run worker processing, Cloud Tasks queueing, Cloud Storage, Vertex AI OCR/embeddings/document analysis, report/PDF generation, and Gmail SMTP for service emails.Uploaded files, OCR text, prompts/instructions, generated outputs, embeddings, processing metadata, operational logs, recipient email addresses for service emails.Managed Vertex AI processing is used to provide the service; Provider does not authorize model training on Customer Content.
StripePayments, subscriptions, invoices, checkout, webhook events, and billing metadata.Billing contact data, payment metadata, subscription status, transaction records, legal-acceptance metadata.Stripe should not receive case files or PHI through normal workflows.
SlackInternal operational notifications, security/error alerts, signup/subscription notices, and workflow status notices if configured.Account metadata, billing/status metadata, redacted error details; Customer Content should not be intentionally sent to Slack.Used only for operational visibility and incident response where enabled.

Provider will require subprocessors to maintain confidentiality and data-protection obligations appropriate to their role.

Provider will provide notice of material subprocessor changes through a website notice, email, customer portal, or other reasonable method. Customer may object to a new subprocessor on reasonable data-protection grounds within the notice period. The parties will work in good faith to resolve the objection.

7. Deletion, Return, And Retention

Customer may delete matters through the product where available or request deletion through Provider's support channel. Provider will delete or return Customer Content according to the agreement, product functionality, and Provider's retention practices.

Unless a different written agreement applies, the current standard policy is that case content is scheduled for deletion 30 days after case creation. Operational logs, billing records, legal-acceptance records, audit records, and security records may be retained for legitimate business, legal, accounting, fraud-prevention, and security purposes. Provider may update retention periods prospectively through the Terms, Privacy Policy, or a written customer agreement.

Backups may retain deleted Customer Content until backup expiry, but backups are protected from ordinary access and are used for continuity, security, and recovery purposes.

8. Security Incident Notice

Provider will notify Customer within five business days after confirming a Security Incident affecting Customer Content, where legally or contractually required. Notice will include information reasonably available to Provider, such as affected systems, affected data categories, known impact, mitigation steps, and any recommended Customer actions.

Security Incident does not include unsuccessful attempts or activity that does not compromise Customer Content, such as blocked scans, pings, rate-limited requests, failed login attempts, or firewall events.

9. Customer Assistance

Provider will provide reasonable assistance, taking into account the nature of the service and information available to Provider, for Customer's deletion, export, security questionnaire, incident response, and regulatory obligations.

Customer remains responsible for managing Authorized Users, configuring access, reviewing generated outputs, maintaining client authorizations, applying minimum-necessary practices, and determining whether Customer's use is legally permitted.

10. Audit And Assurance

Provider may satisfy reasonable security review requests by providing security documentation, the subprocessor list in this DPA, standard questionnaire responses, and relevant vendor assurance materials where available.

Any direct audit requires reasonable written notice, confidentiality, limited scope, no disruption to the service or other customers, and no access to other customers' data. Direct audits may be reserved for enterprise customers, post-incident circumstances, or other cases Provider approves in writing.

11. United States Use Only

The service is intended only for customers and Authorized Users located in the United States and for United States disability-law workflows. Customer may not use the service for non-U.S. customers, non-U.S. Authorized Users, or processing that would require Provider to comply with non-U.S. privacy, data-protection, or international-transfer regimes unless Provider expressly agrees in writing before such use.

Customer is responsible for ensuring that Customer Content submitted to the service is appropriate for this U.S.-only service posture.

12. HIPAA And BAA Terms

This DPA does not by itself make Provider a business associate or create a BAA. If Customer's use requires business associate terms, the parties must execute a separate Business Associate Addendum or HIPAA exhibit, available on request, before Customer submits PHI in a HIPAA-regulated workflow.

Provider's service is designed to support HIPAA-conscious legal workflows when required agreements and customer-side controls are in place.

Appendix 1: Technical And Organizational Measures

Provider's current safeguards include:

  • Encryption in transit for application traffic and backend service calls using HTTPS/TLS or authenticated TLS-protected service connections.
  • Managed encryption at rest for primary cloud data stores and object storage.
  • Supabase Auth for user authentication.
  • Server-side ownership checks for case, file, chat, billing, and report access.
  • Service-role credentials used only in backend/server contexts.
  • Server-issued, time-bound upload and report access patterns instead of public storage locations.
  • Role-based administrative controls and admin session hardening.
  • hCaptcha and rate limiting on authentication-sensitive and high-cost paths.
  • Click-through legal acceptance for the Terms, Privacy Policy, and DPA during onboarding or subscription checkout, with acceptance metadata retained for customer-assurance purposes where available.
  • Security headers including HSTS, CSP, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, and Permissions-Policy.
  • Private source repositories, protected production branch, CodeQL, Dependabot, dependency review, secret scanning, and bundle secret scans.
  • Shared redaction controls for logs and error handling to reduce exposure of tokens, private keys, signed URLs, and raw internal errors.
  • Reviewed migrations and operator-controlled backup procedures before structural production database changes.
  • Incident response, vulnerability intake, and customer assurance materials.

Appendix 2: Processing Details Schedule

This appendix identifies the processing covered by this DPA: what data is processed, whose data may be included, why it is processed, and how long the processing lasts. It is not a separate set of definitions; it is the operational processing schedule for this DPA.

FieldDescription
Subject matterAI-assisted processing of disability-law medical and legal records.
DurationSubscription term plus retention/deletion periods.
Nature of processingUpload, storage, OCR, extraction, embeddings, retrieval, summarization, report generation, chat, support, deletion, security logging, billing.
PurposeProvide, secure, support, troubleshoot, and maintain DisabilityProAI.
Data subjectsLaw-firm users, claimants/clients, medical providers, third parties appearing in records, billing contacts, support contacts.
Data categoriesAccount data, authentication data, medical/legal records, PHI/ePHI where present, case metadata, generated outputs, usage logs, billing metadata, support communications.
Sensitive dataMedical records, disability records, Social Security materials, health information, attorney work product, financial information, government identifiers where present in records.
SubprocessorsListed in Section 6 of this DPA.

Important compliance note

No cloud vendor, AI tool, or software product can make a customer automatically HIPAA compliant. Customers remain responsible for their professional duties, client authorizations, minimum-necessary practices, user management, and any required HIPAA or privacy-law analysis.

Security questions or documentation requests

Contact us for security questions, vulnerability reports, subprocessor details, or Data Processing Agreement review.

security@disabilityproai.com