◇ PROJECT UTAH
TRUST CENTER / UPDATED OCTOBER 24, 2026

Privacy & data

What Utah processes

We process the prompts, source snippets, logs, target information, and other evidence you submit to answer your request or run your authorized research. Google sign-in uses verified identity information to open your account. The account ledger stores hashed identity references, credential hashes, allocation history, usage totals, verified targets, and research-job references.

Where data goes

Cloudflare operates the application and account storage. Utah supplies model access through its managed inference service. Connected app requests go to the public HTTPS API you configure. Cloudflare data reads use your connected account and selected resource. Authorized tool requests are sent to the configured Utah research runner. Google handles Google sign-in, and Stripe handles checkout and payment information. Those providers process data under their own terms. No customer-facing PHI workflow is enabled.

Conversations and local storage

Signed-in browser conversations, drafts, selected skills and activity are saved in account-scoped server storage so you can reopen them. Sign-out clears the active browser view; it does not delete saved server conversations. A previous tab-local draft may be recovered into its matching account. An API key entered into the web app is kept for the browser session, not persistent local storage. Native terminal sessions and transcripts are stored on your computer and can include submitted evidence.

Server records and retention

Saved browser prompts and answers are retained in the account workspace. Connected provider and app API credentials are encrypted on the server and omitted from configuration exports. App responses and Cloudflare contents are displayed for your request; an excerpt is saved in a chat only if you choose to place it in a prompt and send or save that draft. Tool investigations and mission journals retain objectives, results, and evidence needed to show your research history. Account, billing, and investigation records remain until a deletion request is processed, subject to payment, fraud, legal, and backup requirements. We do not promise zero retention by every infrastructure provider. The customer audit trail records account operations without request bodies or credentials. Audit history is trimmed on recorded operations to the latest 2,000 events and 365 days. Framework evidence notes persist with your account. Other operational retention and backup schedules remain assurance-readiness work.

Your choices

Export the current web conversation using Export. New chat starts another saved conversation. Sign out clears the active browser view while retaining your server history. The account audit API exports evidence and audit history; configuration exports omit saved credentials. To request account access, correction, export, or deletion, email michael@blacklabelbots.com from your account identity. We will verify ownership before releasing or deleting account data. Do not send an API key with your request.

Security and changes

Read the security overview for implemented controls and assurance limits. We will update this page as data handling changes. This notice describes the current service; it does not authorize submission of regulated health data.

Signed model-use notice

Before using AI, you must type your name and explicitly agree to the model-use notice. We save that typed name, the authenticated account identifier, the server timestamp, the notice version and a SHA-256 hash of the notice text. You can review the signed receipt from your account menu. Signing events are included in the account audit trail without including the typed name in that trail.