Social comments, privacy and deletion requests

Effective date: November 7, 2026.

Shared Sapience, LLC uses a private Comment Inbox to follow conversations on our posts, organize feedback and help us prepare replies. This notice explains the information involved and how to request access, correction or deletion. It supplements our Privacy Policy.

Where comments come from

The inbox covers YouTube, Instagram, Bluesky, Substack, TikTok and comments on sharedsapience.com. X mentions and replies can be collected through a manual check; automatic X collection is currently off.

We use platform APIs and website/account interfaces, depending on the platform. YouTube, Instagram, Bluesky, X and our own website use API or comment endpoints. Substack uses our publication's logged-in website interfaces and internal endpoints. TikTok uses read-only access to comments shown in our logged-in TikTok Studio account.

This inbox does not collect direct messages or follower lists. A comment can contain personal information that its author chooses to include, so please avoid posting sensitive information.

What we keep and why

Depending on the platform, we keep comment text, display names, handles, profile links, platform identifiers, dates, the relevant post or video link, and thread context. We also keep matching publication context, AI-generated categories and draft replies, operator notes, and records of replies we send or attempt to send.

We use these records to find questions and feedback, understand the discussion and prepare relevant responses. Access is restricted to people and tools working on the publication. Our editors decide whether to send the replies prepared through this workflow.

AI processing and service providers

Automatic sorting uses Ollama's cloud service. It receives the comment, author display name, platform, post title, source links and related publication identifiers, together with fixed sorting instructions.

Comments selected for reply drafting or our Reply Committee can also be processed by Anthropic and OpenAI. Drafting can include the comment and author information, relevant articles or transcripts, publication context, and our reply instructions or saved guidance. Some drafting paths use Claude Code or Codex through individual accounts; their model-improvement controls are turned off. API-based and account-based services have their own retention and safety-processing terms. Ollama states that its cloud prompts and responses are processed transiently and are not used to train models.

Selected feedback has also been saved in private analysis files and version history hosted by GitHub, for use with Claude Code web. New exports exclude YouTube and Instagram comments; older copies remain subject to the limits below. Migadu hosts messages sent to our contact address. Providers may use infrastructure or inference subprocessors and process data in other countries under their own policies.

How long information stays

YouTube records are normally checked against the API daily. A record expires from the active inbox if it has not been successfully verified for 30 days, and comments confirmed removed from YouTube are removed locally. Successful verification renews that period. Unresolved reply-send records and older records with unknown verification dates can be held for owner review, outside ordinary inbox access.

Instagram, Substack, TikTok, Bluesky, X and website-comment records currently have no fixed automatic expiry. They can remain until a deletion request is processed or an owner removes them. Categories, drafts and notes stored with a comment follow that comment's local lifecycle.

Separate Reply Committee session histories, saved reply guidance, older analysis exports and their private version history currently have no fixed expiry. Deletion-request records and minimal identifiers used to prevent re-import are retained separately. Deleting an inbox row does not automatically remove these other copies.

Encrypted backups use a rolling schedule with annual recovery points. Deleted information may remain in a backup for about two years under the current schedule. Backups expire through rotation and are not selectively purged. AI-provider copies follow the relevant provider's terms and available deletion controls.

Request access, correction or deletion

Email [email protected], which is read by Ben Linford. Include the platform, your handle or profile link, the relevant comment or post link, and what you want us to do. We may ask for proportionate proof of control, such as an edit or reply from the original account. Please do not send passwords, access tokens or identity documents.

We will check matching records, coordinate supported removal or correction, and explain any remaining copies, verification limits or records retained to honor the request. Older Instagram records can require additional verification because a matching platform identifier was not captured. Meta deletion callbacks are handled separately; their Facebook app-scoped identifiers are not automatically treated as Instagram commenter identifiers.

Removing our stored copy does not delete the original comment on its platform. Use that platform's controls to remove the original. A comment-data request also leaves newsletter subscriptions and memberships unchanged unless you ask us to address those separately.

Deletion from the active inbox is separate from removal of backups, historical files, version history and provider-held copies. Complete or immediate erasure across every system cannot be guaranteed. Contact [email protected] so we can explain what is possible for your request.