Data Privacy Policy

1. Purpose

This policy describes how Docwright Software LLC ("DocWright," "we," "us") handles data in connection with the DocWright software application. It is intended for institutional customers, IT departments, and accessibility coordinators evaluating DocWright for deployment.

2. How DocWright Processes Data

DocWright is a locally installed desktop application. Documents (DOCX, PPTX, and PDF files) are processed entirely on the end user's own machine. There is no central DocWright server, no shared infrastructure, and no document upload to any DocWright-controlled system at any point during normal use.

2.1 What Never Leaves the Local Machine

• The full content of any document processed by DocWright, including any student, faculty, or institutional data it contains
• File metadata (author, institution name, file paths, etc.)

2.2 What Is Transmitted, and to Whom

Three limited, specific data flows occur:

• License validation and updates: On each application launch, the software sends the license key and a random device identifier (used to count installed seats; no document content) to DocWright's licensing backend, hosted on Vercel, to confirm the license is active and to check for updates. This confirms institutional entitlement only. Signed updates are downloaded from the same backend, and signed updates to the bundled LibreOffice and Python components are downloaded from release assets on DocWright's public GitHub repository, which receives only the ordinary download request. Every update is verified against a cryptographic signature before it is applied.
• Optional AI features: If a user opts in, images are sent directly to the one AI provider configured for that installation — Anthropic (Claude API), Google Gemini, or Azure OpenAI — to generate descriptions or transcriptions. Generating alt text and equation descriptions sends individual cropped images extracted from a document (e.g., a chart, figure, or equation); the surrounding document text, file, and metadata are not included. Transcribing scanned or handwritten pages is a separate, opt-in action that sends the image of one page per request, so everything visible on that page is sent to the provider, and the result is shown to the user for review before it is embedded. These features use an API key supplied and controlled by the institution — DocWright does not provide, proxy, or have visibility into the key, and does not receive the images or results itself. The provider is chosen by the institution but is not centrally locked: a user can change it in Settings.
• Optional issue reporting: If a user chooses to report a problem via the in-app "Report an Issue" link, a Google Form opens, pre-filled with only the processed file's name and a short summary of issues found and fixes applied — never document content. The submitting user manually enters their own name, department, and contact email if they choose to submit. There is no file-attachment capability on this form.

3. Data Retention

DocWright does not retain document content, student data, or institutional data on any DocWright-controlled system, because such data is never transmitted to DocWright in the first place. The only data Docwright Software LLC retains centrally is administrative license information: institution name, license key, tier, term, and contact details necessary to administer the license relationship, together with seat registrations (a random device identifier recorded against each license key to count installed seats).

4. Third-Party Sub-Processors

DocWright relies on the following third-party services in connection with the limited data flows described above:

• Vercel — hosts the license validation and update backend (license key and a random device identifier; no document data).
• Upstash — managed Redis database used to record seat registrations (license key and random device identifier only; no document data). Primary region: AWS us-east-1 (N. Virginia, USA).
• GitHub — hosts signed update files for the bundled LibreOffice and Python components. Receives only ordinary download requests; no document data.
• AI provider chosen by the institution (Anthropic, Google Gemini, or Azure OpenAI) — used only for the optional AI features, under the institution's own API key and its own agreement with the provider; DocWright is not a party to that agreement. Under Anthropic's commercial API terms and Microsoft's Azure OpenAI terms, data is not used to train the provider's models. Under Google's Gemini API terms the same is true for paid services, but for unpaid (free-tier) use Google may use submitted content to improve its products and human reviewers may read it, so institutions that select Gemini should use paid-tier keys. Azure OpenAI retains prompts and outputs for up to 30 days for abuse monitoring unless modified abuse monitoring is approved for the institution's subscription. DocWright recommends institutions review their chosen provider's commercial terms directly, particularly regarding data retention, for documents that may contain student work or personally identifiable information, and consider negotiating a Zero Data Retention arrangement (Anthropic) or modified abuse monitoring (Microsoft) if a stronger contractual guarantee is desired.
• Google (Forms) — used only if a user opts to submit feedback via the optional "Report an Issue" link. Receives only a file name and processing summary (never document content), plus whatever the submitting user manually enters.

5. FERPA and Institutional Data Protection

Because document content is never transmitted to or stored by DocWright, and the optional AI features operate under the institution's own account with its chosen AI provider, DocWright's architecture is designed to minimize exposure under FERPA and similar data protection requirements. The AI features can be disabled by the institution by not providing an API key; all other accessibility remediation runs locally with no document-related network transmission. DocWright does not currently enforce this centrally, so a user who enters their own key in Settings can enable the AI features; institutions that need to prevent this should address it in their acceptable-use policy.

6. Security of License and Configuration Data

The license key and (if configured) the institution's AI provider API key (Anthropic, Google Gemini, or Azure OpenAI) are stored locally in a configuration file on each installed machine, encrypted at rest using the operating system's built-in data protection (Windows DPAPI, scoped to the logged-in user account). License, update, and AI-feature requests are transmitted via HTTPS. Institutions are responsible for the physical and account-level security of machines on which DocWright is installed, consistent with their own device management policies.

7. Changes to This Policy

Docwright Software LLC may update this policy as the product or company evolves. Material changes affecting how institutional or student data is handled will be communicated to licensed institutions directly.

8. Contact

Questions about this policy or DocWright's data practices may be directed to: