Privacy — Regard

Reading without an account

Reading Regard needs no account. Anonymous reading sets no cookies, runs no analytics, loads no third-party scripts, and embeds no trackers or advertising. Fonts are bundled locally. External links carry rel="noopener". Choosing Tamil or English for the interface stores one preference cookie (regard-ui-lang) on your own device — nothing is sent anywhere.

When you create an account

Accounts exist so you can write, save drafts and publish. Stored:

Account information: email address and authentication data, held by our auth provider (Supabase Auth) — Regard never sees your password.
Public profile: username, display name, bio, website and avatar if you add them. Usernames and display names are public by design.
Publication data: everything you publish (title, body, topics, dates) is public and appears in search, feeds and sitemaps.
Draft data: drafts, pending and hidden pieces are private — visible only to you (and to moderators when flagged for review, who see only what review requires).
Operational logs: basic server logs needed to run the service. No behavioural profiling, no adtech.
Reports: if you report a piece, the report (category, detail, time) is kept for moderation. Anonymous reports keep no identity.

Your writing data is never sold. Analytics, if ever enabled, will be privacy-preserving (self-hosted or aggregate-only) and documented here before it ships. See docs/ANALYTICS.md in the repository for the current state: none.

Deletion and export

Your writing is portable: export everything any time from Settings (Markdown or JSON). Deleting your account removes your profile and all of your writing immediately; published URLs stop resolving. Backups, if any, age out rather than being mined. Drafts never leave your account except through export or publication by you.

Writing about a remembered life

Regard’s case studies draw on long conversation histories between a person and AI systems. Persistent memory makes those systems more useful — less re-explaining, more continuity — and the same mechanism means a great deal of intimate context accumulates in one place.

Deep personalization creates value precisely because the system knows more about the person. The same fact makes governance, consent and control more important, not less.

Regard does not claim knowledge of any AI provider’s internal data practices beyond what is publicly documented, and it does not allege data sale, training use, retention or surveillance by any named company. Its argument is architectural: a system whose usefulness scales with how much it holds about a person will tend toward accumulating that context, so the mechanisms that govern it — storage, retention, access, real deletion, export, the ability to leave — matter more as the personalization deepens.

Sensitive-content policy

The fact that something appears in a conversation history is not a licence to publish it. The following categories are excluded from public Regard writing unless the author has specifically approved a particular passage, recorded as a review gate in that piece’s metadata. This policy is kept as a single reusable configuration so every future piece inherits it.

Medical history and diagnoses: No medication history, substance-use history, diagnoses, or accounts of specific medical episodes — the author’s or anyone else’s.
Sexual history: No sexual history or sexual detail about the author or any identifiable person.
Private individuals by name: No names of private romantic partners, family members, colleagues or officials. Others appear as a role or a single-letter pseudonym, never a name.
Third-party private histories: No private history, medical information or correspondence belonging to family members or other third parties.
Contact and location data: No phone numbers, home or work addresses, precise locations, or IP addresses.
Financial detail: No account balances, exact figures, or identifiable financial records. Broad statements about income pressure or subscription cost are allowed.
Credentials and secrets: No passwords, tokens, API keys, or anything that functions as a credential.
Unpublished employment disputes: No unpublished workplace disputes that name or identify individual officers or colleagues. General statements about occupational targets, hierarchy and pressure are allowed.
Private correspondence: No reproduction of private messages or emails from identifiable people.

What may be discussed

Within those limits, and in a non-exploitative way, Regard writing may address:

work frustration and career pressure
financial limitations, in general terms
attachment, ambition, loneliness
philosophy and intellectual interests
software projects and AI usage
broad biographical transitions