Telemetry and Browser Storage Notice

1. Different records serve different purposes

SourceWarden LLC operates EvidenceGuard. This Notice supplements the Privacy Policy and distinguishes necessary service operation, browser fault reporting, optional product measurement, customer audit history and research. Calling a record “telemetry” does not make it anonymous, optional or exempt from privacy law. Terms acceptance is not permission for every one of these activities.

Contact: kun11451454188@gmail.com, subject “EvidenceGuard privacy.” Postal contact: SourceWarden LLC, 7155 Kindred Street, Philadelphia, PA 19149-1132, United States.

2. Browser storage used for requested functions

Authentication state. The authentication SDK can store a session and refresh state to maintain a requested sign-in. The exact session lifecycle depends on the configured authentication service. Persistent browser storage can survive closing a tab; it is not automatically a short-lived cookie. Signing out or clearing storage can remove local access without deleting the underlying account or every provider record.

Organization selection. The application uses local storage, including `evidenceguard.organization-id`, to remember the selected organization. The application or browser may update or remove it. It is not an advertising identity and must not be repurposed to track unrelated sites.

Sign-in return destination. Session storage, including `evidenceguard.auth-return-to`, supports returning to an allowed destination after authentication. It is not permission to record a user’s complete navigation history. Authentication return details must not be copied into ordinary analytics.

In-memory operational state. Temporary state can deduplicate fault reporting, limit frequency and support the current page. It is not the same as a persistent cross-site identifier.

Third-party sign-in pages and providers can use their own storage under their applicable notices. We do not describe the Service as “cookie-free” or claim that the small first-party inventory exhausts every provider’s browser behavior. Blocking necessary storage may prevent a requested function.

3. Limited browser fault reporting

In an authenticated organization context, the first-party reporter can send a constrained error category, coarse application area, status code and request reference to our same-origin diagnostic endpoint. It is designed not to include document text, filenames, full customer-specific routes, exception messages or stacks, screenshots, keystrokes, DOM snapshots or session replay in the diagnostic event payload.

Authentication and organization context are still used to authorize that request, and service records may permit correlation. “Content-free diagnostic payload” does not mean “no personal data anywhere.” Reporting is limited and deduplicated; absence of a reported event is not proof that no failure occurred.

4. Server and infrastructure records

Application and infrastructure layers have separate records. Application diagnostics can include timestamp, method, route pattern, response status, duration, request reference and constrained exception type. Proxy, web-server, authentication, hosting or network records may additionally contain IP-related information, request paths, browser information and referrers. They are not described as identical to the minimized browser payload.

We use those records for operating, securing and diagnosing the Service. They are not a license to log confidential documents, working signed links, complete authentication headers or credentials. Routine logs under our control follow the Privacy Policy’s 30-day maximum; specifically isolated incident records and provider-controlled records follow their separately stated rules. Size-based log rotation alone is not proof of a calendar retention limit.

5. Optional product statistics

The organization product-statistics policy defaults to off. When applicable conditions permit it, a contribution may record an organization, day, feature, daily user-derived identifier and 30-day expiry. Stored contribution rows do not include source text, source location or a document asset ID, although an asset reference may be sent to check access and eligibility before storage.

The implemented mechanism permits at most one stored contribution per person, feature and day, excludes restricted or unknown content classifications, uses nonexpired prior-day data for reporting, suppresses groups below five contributors and rounds reported counts. Those safeguards reduce detail; underlying records remain pseudonymous, not necessarily anonymous.

Some current feature paths can send an eligibility request even where the server does not store a contribution. An HTTP 204 response is not proof of collection or noncollection. Disabling the organization setting stops eligibility for new storage under that policy, not every service request, prior record, security log or audit event.

An organization administrator cannot supply a legally required individual choice simply by enabling a switch. Where such a choice or additional device-consent control is required but not available, the affected optional measurement must remain disabled. No control should claim to prevent transmissions it does not prevent.

6. Audit, quality and research boundaries

Customer audit records explain who acted, when, within which organization and source-version context. They can include personal identifiers necessary for that function and are not governed by a marketing-analytics switch. Retaining minimal audit context does not justify retaining deleted document passages indefinitely.

Operational counts of errors, timing or resource use are different from a content-bearing quality dataset. Quality testing may use synthetic, appropriately licensed public or specifically authorized evaluation material. Customer corrections and reviewer decisions are not automatically training data or ground truth. This baseline does not offer customer-content research or general-purpose model training.

7. Choices, signals and contact

Optional analytics, marketing, research and external-AI authorizations are separate choices. We do not offer advertising-pixel or session-replay collection under this version. A privacy refusal must not be silently converted to acceptance, and a failed preference save does not authorize processing dependent on that choice.

We do not use a “Do Not Track” signal to switch off necessary service processing. Our no-sale/no-cross-context-sharing rule applies with or without a Global Privacy Control signal. Where law requires action on a recognized opt-out signal, we honor that requirement for the affected activity. No statement here implies that a browser signal deletes server content or disables authentication.

Request access, correction, objection or deletion through the Privacy Policy’s contact and procedure. Clearing local browser storage and requesting server-side deletion have different effects. Retention, recipients and international handling follow the Privacy Policy and applicable DPA.