Shepherd Privacy Policy

1. Overview

Shepherd helps Discord communities offer guided support in difficult moments. This Privacy Policy explains what information Shepherd may process, why it is used, and how to contact us with questions.

2. Information Shepherd may process

Depending on server configuration and user interaction, Shepherd may process Discord server IDs, channel IDs, user IDs, usernames or display names, command usage, support choices, support progress, language preferences, analytics events, scan settings, intervention records, classification results, suppression reasons, and message content or metadata from channels where admins enable scanning.

Shepherd may also process attachment metadata, link metadata, and short contextual excerpts when needed to classify risk, explain why support was offered, debug the service, or improve safety behavior.

If you create a Shepherd account, Shepherd processes the identity your sign-in provider returns (such as a Discord, Twitch, or Google account identifier, display name, avatar URL, and where provided an email address), the platform connections you link, and your role.

If your account uses engine credits, Shepherd keeps a credit ledger: your balance, each grant, each hold placed when a run starts, and each charge settled when it finishes, with the run it belongs to. This is an accounting record of what was spent, not a copy of what you submitted. Shepherd does not process card numbers or payment credentials, because credits are not sold today and there is no payment processor.

When you use an engine, Shepherd processes what you submit (for example a claim, pasted text, or a link), the output produced for it, and request metadata such as timing, model and token usage, engine version, and error and security logs. Anyone who can sign in to anapproved account, or use an API credential issued for it, can see that account's engine history and ledger — so treat account access as you would a payment method. An account that is signed in but not yet approved, or whose access has been suspended, cannot read that history from the site: the record still belongs to it and is not deleted, but you have to ask for it, and Shepherd will provide it.

One detail worth stating rather than leaving you to discover it: when two accounts submit the identical request, Shepherd does the work once and both receive the same job reference. An approved account holding a job reference can read that job’s result, so a result can be visible to another approved account that asked the same question. It is the same answer to the same question, never a window into anything else that account did, and the reference itself is an unguessable identifier Shepherd only ever returns to a submitter.

3. How information is used

We use information to operate Shepherd, provide guided support, classify configured chat context, create private support spaces, prevent repeated or unwanted interventions, maintain safety boundaries, debug issues, measure aggregate outcomes, and improve the product.

4. What Shepherd does not do

Shepherd does not sell personal data. Public analytics do not expose private conversations, raw channel messages, user-level records, or identifiable server data. Faith-based support is optional and based on user choice.

5. Service providers

Shepherd may use service providers to host the website and bot, store data, classify messages, send Discord responses, and measure aggregate analytics. These providers may include Discord, Vercel, Railway, Supabase, OpenAI, and PostHog or similar infrastructure and analytics providers.

Model providers. Running an engine sends the content being assessed, and the instructions for assessing it, to a model provider (currently OpenAI). On Shepherd’s provider account every data-sharing and training option is turned off, API call logging is disabled across the organisation, and each request is sent asking the provider not to store it. Shepherd does no fine-tuning, so nothing you send trains a model.

One limit, stated plainly rather than left to the word “briefly”: providers may still retain a request for a short period for their own abuse-monitoring and reliability purposes, under their terms and outside Shepherd’s control. Shepherd has not yet obtained a zero-retention arrangement that would remove that window. If that changes, this page changes with it.

6. YouTube API Services

Shepherd uses YouTube API Services to find and present official music from YouTube. Shepherd's use of information received from YouTube API Services complies with the YouTube Terms of Service, and Google's handling of data is described in the Google Privacy Policy.

Shepherd accesses only public YouTube data (video and channel metadata such as titles, video IDs, embeddability, durations, and channel or subscriber information) solely to resolve and verify the official source of a song. Shepherd does not access, request, or store any personal YouTube or Google account data, and does not require users to sign in with a Google or YouTube account.

Any public YouTube metadata Shepherd stores is used only to serve official content, is refreshed periodically, and is deleted on request at hello@sovereignify.ai.

7. Retention and deletion

We keep information only as long as reasonably needed to operate, secure, debug, and improve Shepherd, unless a longer period is required for legitimate safety, legal, or operational reasons. Server administrators or users may contact us to request deletion or review of information associated with Shepherd.

Submitted content and accounting records are kept on different clocks. What you submit to an engine, and the output produced for it, are kept for as long as needed to deliver and support that result and can be deleted on request. Credit ledger entries are accounting records: they are retained as a record of what was granted and spent, and are not deleted on request while they are needed for that purpose. A deletion request can remove submitted content without erasing the record that a charge occurred.

8. If you submit other people's information

If you send Shepherd content about someone else — including through an engine or a programmatic credential — you are responsible for having the right to share it, and for not sending more than the task needs. Do not submit another person's support conversation, private messages, or identifying details unless you are permitted to. If you are using Shepherd on behalf of an organisation and expect to send personal data routinely, contact us before you do: that arrangement needs a written data-processing agreement, which is not offered through self-service today.

9. Children and sensitive situations

Shepherd may be used in communities that include minors. Shepherd is not a replacement for parents, guardians, trusted adults, emergency services, or qualified professionals. In serious or urgent situations, users should contact local emergency services, a crisis hotline, or a trusted person immediately.

10. Your choices

Server administrators can remove Shepherd from a server or change which channels are configured. Users can decline support prompts or stop interacting with Shepherd. For privacy questions or deletion requests, contact us at hello@sovereignify.ai.

11. Changes

We may update this Privacy Policy as Shepherd changes. The updated date above shows when this page last changed.