Privacy policy · Draft
Privacy policy
Draft pending owner approval. This text describes what the app and this website do today. It is not yet in force and has not been checked by a lawyer.
What Sekmet keeps, why, who else touches it, and how to make it stop.
Who we are
Sekmet is a training and nutrition app, plus this website. The controller of your data under the GDPR is Pavloff Corporation (entrepreneur individuel, EI), SIREN 921 265 104, France. Contact: [email protected].
What this covers
- The Sekmet app.
- This website, including the Sekmet sign-in and consent page used when you connect an assistant such as ChatGPT.
- The optional connection that lets ChatGPT read a snapshot of your data. Connections are limited to people we have invited.
What we process
| Data | What it is | Where it lives |
|---|---|---|
| Account | Email address, optional display name, and your sign-in credentials. | Supabase (authentication). |
| Synced app data | Workouts, recovery entries, goals, nutrition entries and synced profile records that you store in the app while signed in. | Supabase database, tied to your account. |
| Device-only data | Stretching and isometrics plans and history, reminder preferences, and your training profile (goals, experience, availability, equipment, limits, units, planner note and, if you choose, birth year, sex and height). Android Health Connect data is read for display on the device. | Your device only. Not uploaded, except the sport-context copy you choose to share (see below). |
| Meal photos | Meal photo analysis is not available to users today. No photo is analysed or sent to an AI provider on your behalf. | Not applicable. |
| Usage analytics | Pseudonymous usage events without content, such as that a workout was completed or a sync failed, sent only by app builds configured with a PostHog key. Each event carries its name, a few fixed categorical or yes/no values, the source “mobile-app”, and an identifier: your Sekmet account identifier when you are signed in, otherwise an identifier generated on your device. Your IP address reaches PostHog with each request. No names, plan or routine names, notes, health values, email address or free text are sent, and anything not on a fixed allowlist is dropped before sending. See “Usage analytics and diagnostics” below. | PostHog, on its US host. |
| Crash reports and diagnostics | Sent only by app builds configured with a Sentry DSN. Crash and error reports: error details and stack trace, recent app events leading up to the error, and device and app details (such as model, operating system and app version). Performance traces, such as app start timings, from about 10% of sessions. A release-health session record at each app launch: app release, environment, device and operating system. Your IP address reaches Sentry with each request. See “Usage analytics and diagnostics” below. | Sentry. |
| Connection requests | The request ChatGPT makes (client name, client identifier, return address, requested scope), and what you approve or deny. | Supabase. |
| Sharing snapshots | Only the categories you tick on the consent screen, copied at the moment you approve (see below). | Supabase, until they expire or you disconnect. |
| Access history | For each event: a category, a result (read, denied, approved or revoked) and a time. No values, no search text, no addresses, no names. | Supabase. |
| Browser session | Your Sekmet sign-in session, kept in this browser’s local storage after you sign in on the consent page. | Your browser. |
| Website visits | No cookies, no analytics, no advertising and no third-party requests (fonts are served from this site). Our host necessarily handles your IP address and request details to deliver pages. | Cloudflare. |
The ChatGPT connection, in detail
Nothing is shared until you approve it on the consent screen. You choose among five categories:
- Sport context (ticked by default): your confirmed goals, experience level, training days, session length, equipment and whether you train at home. It comes from the confirmed fields of the training profile stored in the browser where you approve, and is uploaded as part of the snapshot only if you tick this box.
- Active plans (ticked by default): your active training plans, with the names you gave them and their structured targets.
- 8-week summaries (ticked by default): aggregated training figures for the last eight weeks.
- Sessions and sets (8 weeks) (not ticked): individual sessions and sets.
- Body measurements (8 weeks) (not ticked): measurements you logged.
The consent screen shows the exact values before you decide. By design these are never included: medical fields, profile and session notes, saved GPS routes, and demographic fields (birth year, sex, height). Names you typed into plans and exercises are included as text, so read them before you share.
What happens next:
- The snapshot is fixed at the moment you approve. New or changed data is not added. The history window is the 56 days before approval.
- Access lasts 24 hours from that moment, then ends on its own. Approving again replaces the earlier grant and deletes its snapshot.
- ChatGPT can search and read only the categories you approved. Each read is recorded in your access history. Search text and values are never recorded.
- A request you do not approve within ten minutes expires, and its preview is discarded.
- If plan drafts are enabled for your account (they are off by default), ChatGPT can submit a draft plan. A draft is only a suggestion held for you to review in the app. It changes no plan and no training record, and it expires after seven days.
The data you allow is sent to ChatGPT and is then handled under your ChatGPT account’s terms and data policies, not ours. We cannot see or control what it does with it, and disconnecting does not erase copies it already received.
Health and fitness data
Training, body and recovery information can reveal things about your health, so we treat all of it as sensitive. Health Connect data stays on your device. We do not sell your health and fitness data, use it for advertising, or use it to train AI models. Sekmet is not a medical service and does not give medical advice.
Usage analytics and diagnostics
Only app builds configured with a PostHog key send the usage events described above, and only app builds configured with a Sentry DSN send crash reports and diagnostics to Sentry. In those builds both are on by default.
Usage events are pseudonymous: when you are signed in, the identifier is your Sekmet account identifier, so events can be linked to your account. Each event carries only its name, fixed categorical or yes/no values, the source “mobile-app” and that identifier, and your IP address reaches PostHog with each request. Events go to PostHog’s US host, so this is a transfer of data outside the European Union.
Sentry receives crash and error reports when the app hits an error, performance traces from about 10% of sessions, and a release-health session record at each app launch (app release, environment, device and operating system), even when nothing goes wrong. Your IP address reaches Sentry with each request. Depending on the Sentry project a build uses, this data may be processed outside the European Union.
The app has no switch to turn analytics or diagnostics off. To object, or to have events or reports already sent removed, write to [email protected]. Deleting your account deletes our own records but does not by itself erase what PostHog or Sentry already hold.
Who else handles data
- Supabase: database, sign-in and server functions.
- Cloudflare: hosting, content delivery and DNS for this website.
- PostHog: receives the pseudonymous usage events described above, only from app builds configured with a PostHog key.
- Sentry: receives the crash reports, performance traces and session records described above, only from app builds configured with a Sentry DSN.
- OpenAI (ChatGPT): a recipient only when you connect it and approve a category. It is not our processor.
Apart from PostHog and Sentry as described above, the app uses no analytics, advertising or tracking services, and this website uses none at all. If meal photo analysis is opened to users, this section will name its provider first.
Legal bases (proposal for the owner)
This section is a proposal pending the owner’s decision.
- Your account, synced data and the ChatGPT connection: needed to provide the service you asked for (contract, GDPR Article 6(1)(b)).
- Health and fitness data: your explicit consent (GDPR Article 9(2)(a)). You can withdraw it by deleting the data, disconnecting ChatGPT or deleting your account.
- Usage analytics and diagnostics: our legitimate interest in keeping the app working and improving it (GDPR Article 6(1)(f)). These are on by default in builds that send them, and EU data protection authorities often contest legitimate interest as a basis for on-by-default analytics.
How long we keep it
| Account and synced data | Until you delete your account. |
| Sharing snapshots | Until they expire (24 hours) or you disconnect, whichever is first. Cleanup runs hourly. |
| Access history | 90 days. The app shows the latest 100 events. |
| Expired grant records | Removed 90 days after they expire. |
| Draft plans (if enabled) | Seven days, then cleared. A short record of the request identifier is kept for 30 days to prevent duplicates. |
Your choices
- Stop sharing: in the app, Profile & Settings → Preferences → Connections → ChatGPT → Disconnect ChatGPT. It revokes every grant and deletes its snapshots.
- Delete your account: in the app. Your account, its synced records and its sharing and history records are deleted on the server, then local data is cleared.
- Access, export, correction or deletion: use the app where it offers the action, or write to [email protected]. We answer within 30 days.
- Anything else: write to the same address.
Under the GDPR you also have the right to object to or ask Pavloff Corporation to restrict processing, and to receive your data in a portable form. If you think Pavloff Corporation has not handled your data properly, you can complain to the CNIL, the French data protection authority (cnil.fr), or to the authority in your own country.
Security
Each account’s data is separated by database row-level security. An assistant’s access token cannot reach your account tables, only the snapshot you approved. No system is perfectly secure.
Children
Sekmet is for people aged 16 or over. We do not knowingly collect data from anyone younger. If you believe a younger person has an account, write to [email protected] and we will delete it.
Changes
We will change the date below when this text changes. Last updated: set on approval.