Privacy
What this platform does and does not do with your data. This describes the actual implementation -- nothing here is a legal promise beyond what the code does.
No accounts, no tracking of who you are
This platform has no account system. There is nothing to sign up for, and no personal information is collected about who is using it. Every visit is anonymous.
Anonymous, aggregate site analytics
The production deployment uses Vercel Web Analytics and Speed Insights to see, in aggregate, which pages are visited and how quickly the site loads for real users. This is limited to anonymous page-view counts and performance timing -- there are no custom events, no cookies, and no per-person profile. It never records what you search for, what you upload, what you type into Advocate or Copilot, or any other content you enter. These tools run only on the production deployment; local development and any non-Vercel host see no network calls from them at all.
Advocate workspaces are stored only in your browser
Anything you build in Advocate -- a selected place, chosen evidence, notes, or a generated brief -- is saved to your browser's local storage (IndexedDB) on your own device. It is never sent to or stored on our server. Clearing your browser's site data for this platform will delete it; use the "Export JSON" button first if you want to keep a copy. Workspaces do not sync between devices or browsers.
Uploaded documents
A document you upload for analysis (an agenda, staff report, or similar) is processed in memory, on our server, for that single request only. It is never written to disk and never stored after the response is returned. Only the extracted findings (detected topics, places mentioned, and similar) -- never the original file -- are saved to your local workspace. Do not upload anything containing protected health information (PHI); this platform has no PHI storage or handling capability. See the technical documentation for the full detail.
Optional AI-assisted Copilot mode
Copilot works fully without any AI provider by default (deterministic mode). If this deployment's operator has configured an AI provider and you specifically choose to use AI-assisted mode for a given question, the evidence and text relevant to that one request are sent to that provider (currently Anthropic). This never happens automatically, and never for document analysis or the rest of the Advocate workflow.
What we log
Server request logs record the request path, method, response status, and timing -- never the content of a request, an uploaded document, or your workspace data.
Public health data only
Every metric and score this platform shows comes from public federal, state, county, or transit-agency sources -- see the Data page for the full source list. Nothing here identifies or is derived from any individual person.