Privacy

Uploaded research and AI processing

Your uploaded files and authoring work are stored with Supabase and access within Subset is restricted to your organisation. Your data is analysed securely to compute findings. Figures are always calculated from your data, never written by AI. Publishing shares approved findings through the experience's link, not raw response rows.

Subset uses the Lovable AI Gateway to send research inputs to OpenAI models for analysis proposals and written interpretation. Depending on the size of the file, these inputs include full structured response data or a structured summary containing column metadata, computed aggregates and sampled response rows. Interpretation also uses calculated findings and research context. This processing is not limited to column names.

Subset's creation assistant sees only your column names and your words. This narrower boundary applies to the creation assistant, not to the separate analysis of your uploaded data.

We do not make a blanket promise that AI providers never retain research inputs or use them to train models. Provider handling depends on the service and terms that apply to the processing. Lovable's privacy policy and OpenAI's privacy policy describe their data handling.

Account and browser storage

Supabase processes your email address, account details and sign-in requests. Subset stores your sign-in session in your browser so you stay signed in, and stores local preferences and an advisory template choice. If you choose Google sign-in when it is available, Google also processes that sign-in. The site loads fonts from Google Fonts, which receives the requests needed to deliver them.

Site visits

Lovable collects site analytics for us. For every visitor, it records pages visited, where visitors came from, country, browser, operating system and device type, time on site, page loading speed, errors and clicks. This analytics uses no cookies, stores nothing on the visitor's device, does not keep IP addresses and does not identify visitors. Detailed data is kept for 90 days and daily summaries for a year. Lovable's privacy policy explains how it handles this data.

Reading a published experience

When you read an experience published with Subset, we record anonymous reading events, such as which reading length you chose, which sections you reached and how long the experience was open. We do not use cookies or similar technologies for this, we do not store your IP address, and we do not link your reading across experiences or visits. These events are stored with Supabase. Workspace access is limited to anonymous totals; the author-facing engagement view is not yet available.

Editing proposed text

When you edit text that Subset proposed, we keep both versions so your workspace can see how its experiences were refined. This edit-history record is stored with Supabase and is private to your organisation. AI processing of research inputs is described separately above.

Retention

Within Subset, raw reading events and AI call logs, including recorded prompts and responses, are retained for 13 months, then removed by a daily job. Anonymous reading counts remain. Edit history remains for the life of the experience and is deleted with it. Computed figures are omitted from edit logs. These periods describe Subset's records, not an AI provider's retention.