Privacy Policy

When you use Your Block, we collect:

• Account information: email address, name, and account identifiers (via Clerk authentication).
• Submissions: resident questions, concerns, and answers you write or approve on the platform.
• Activity logs: records of actions taken (approvals, publishing, status changes) with timestamps.
• Postal or zip code: if you enter one at onboarding to find the project for your area. It is used for that lookup only and is not stored.
• Invitations: when a consultant invites a colleague to a project team, the email address they enter is kept on the invitation record, shown to that project’s consultants, and noted in the activity log, until the invitation is used, expires, or is revoked.

Your Block uses Clerk for authentication. Your email, name, and unique Clerk user ID are stored by Clerk and referenced in Your Block’s database. Clerk is a third-party identity provider. For details about how Clerk handles your data, see the Clerk Privacy Policy.

When you submit a concern or question on Your Block, the text is stored in the platform database (Neon PostgreSQL) and associated with your account. By default, submissions are private and visible only to you and the consultant team. If a consultant publishes your thread to the community, other registered residents can then read it.

When drafts are generated, or when a project has turned on instant answers or automatic answers:

• Embeddings (Voyage AI): Resident concern text is sent to Voyage to generate a vector embedding for similarity matching against previously approved answers and against the project’s documents. Voyage processes your text to produce a mathematical representation.
• Drafting for the project team (Anthropic Claude): When a consultant generates a draft answer, the resident’s question and project knowledge-base excerpts are sent to Anthropic’s API (using the Claude model) to generate suggested text. A draft on this path is shown to a resident only after a consultant has approved it.
• Instant answers (Anthropic Claude): When instant answers are enabled for a project and a resident’s new question closely matches a previously approved answer, the question and that approved answer are sent to Anthropic automatically, to check whether the approved answer actually addresses the question; only a "yes" allows the approved text to be shown. The text shown is the consultant-approved answer, verbatim, with its source.
• Automatic answers from the project documents (Anthropic Claude): When a project has turned this on and no instant answer applied, the resident’s question and passages retrieved from the project’s documents are sent to Anthropic to write an answer from those passages, and sent again to check that every claim in the answer is supported by them. If that check passes and the answer meets the project’s confidence bar, the AI-written answer is shown to the resident immediately, without a consultant reading it first. It is labelled as an automatic answer, lists the document passages it was based on, and carries a button to send the question to a person instead. The project team is notified at once and can retract the answer at any time. A project can turn this on only after an evaluation of its own documents recorded no wrong answers, and it turns itself off when the documents or the AI models change until a new evaluation is recorded.
• Anthropic may log usage for system monitoring and abuse prevention (see the Anthropic Privacy Policy).

These calls are made only when:

• A consultant explicitly clicks "Generate draft".
• Instant answers or automatic answers are enabled for the project and a resident submits a new concern (embedding on every submission; the Anthropic instant-answer check only when a close match is found; the Anthropic writing and grounding calls only when automatic answers are on and no instant answer applied).
• A consultant starts an evaluation of the project’s documents. That process sends document passages to Anthropic to write and judge test questions; it does not use resident text.

You should be aware that your concern text is shared with these external services when these features are used, and that on a project with automatic answers turned on, a response written by an AI model may be sent to you without a person reading it first. Consult the privacy policies of Voyage AI and Anthropic for details on how they handle your data.

Your Block uses Resend (an email service provider) to send notifications to consultants and residents. Email notifications include:

• The concern or answer title.
• A link to the platform.

The full concern text is never included in an email to protect privacy. Invitation emails to people a consultant invites carry the project’s name and a one-time link to join its team. Resend processes email addresses and delivery information. See the Resend Privacy Policy for details.

Your Block records audit events for every state change: approvals, publishing, status transitions, and membership changes. Each audit event includes:

• Who took the action (consultant or resident).
• What changed (before and after values).
• When the action occurred.

These records are retained indefinitely to ensure accountability and transparency. Audit logs are not deleted even if the related concern or answer is removed.

Your account data and submissions are retained for the life of the project. There is currently no self-service deletion; requests should go to Mandate support, and audit records and published threads may be retained for legal or regulatory reasons.

You can access your own submission history by logging in. Residents cannot see other residents’ submissions unless the consultant chooses to publish them. Consultants can see all submissions in their project and control whether a thread is published to the registered community; a published thread can be made private again. Submitted text itself is not edited or deleted by consultants.

Your Block is hosted on Vercel and uses industry-standard encryption for data in transit (HTTPS). When something fails on the server, a short technical error notice (the failing step and the kind of error, never your submissions, your email address or your IP address) is sent to Sentry so the operator can respond. The database is encrypted and access is restricted to authorized personnel and application code. However, no system is completely secure, and we cannot guarantee absolute protection against unauthorized access.

This privacy policy may be updated. Continued use of Your Block after changes indicates your acceptance of the updated policy.

If you have questions about this privacy policy or how your data is handled, contact the consultant team or Mandate support.