Privacy policy
What we collect
Account and workspace details; the briefs you write and the assets you upload; the deliverables produced; and run records: what was understood, which models ran, which checks passed, what was refused, and what it cost. We also collect ordinary product telemetry: pages, errors, and performance.
Why we collect it
To run the work you ask for, to price and bill it, to show you how a result was made, to enforce our policies, and to improve the product's reliability. We rely on the contract with you as our legal basis for the first four, and legitimate interest for the last.
Training
Your briefs, uploads and deliverables are never used to train or fine-tune a model, ours or a provider's. Where a provider's default terms would permit training, we contract out of it before routing any work to them; a provider that will not agree is not on the roster.
Who processes it
Generative and judgement model providers, cloud hosting, payment processing, error monitoring and email delivery. Each is named in the sub-processor disclosure with the data it touches and the region it operates in. We give 30 days' notice before adding one.
Retention
Deliverables you keep are held until you delete them. Rejected outputs clear after 30 days. Run records and activity are kept for 12 months, uploaded assets until you delete them, and refusals with their reasons for 24 months. Deleting a production leaves a tombstone: what it was, who deleted it, and what it cost.
Your rights
Access, correction, export, deletion, restriction and objection, exercised from the product where possible and by writing to [email protected] otherwise. We answer within 30 days. If you are in the EEA or UK, our data processing addendum is incorporated into your terms on request.
Transfers and security
Data is processed in the EU and the US under standard contractual clauses. Storage is encrypted at rest, transport is TLS 1.3, and access by our staff is role-based, logged, and limited to support and abuse investigation.