Privacy and Data Handling

Text submitted for editing is processed in memory for the active request and returned to your browser. The platform does not store the original passage or the edited output in its application database, does not write either to application-authored logs, and does not build a copy, archive, index or training set from them. Closing or refreshing the page discards the result.

Only the minimum required to operate an account:

• Account identity handled by the authentication system (email address and password hash).
• Plan, subscription status, billing-cycle dates and entitlement status.
• Numeric usage counters: the number of processed words, never the words themselves.
• Short-lived, non-reversible request fingerprints used for rate limiting and duplicate-submission protection.
• Support messages you choose to send through the platform.

Editing is carried out by a securely configured server-side process. Model credentials are held on the server and are never exposed to the browser. Passages are transmitted for the single request and are not retained by the platform afterwards.

The services involved, and what each receives:

• Lovable Cloud (hosting, database and authentication): account, plan, numeric usage and support records. Passage text passes through the server for the request only and is not written to the application database or application-authored logs.
• Lovable AI Gateway, which forwards each request to the underlying model providers (currently OpenAI and Google): the passage and editing instructions for that request. Lev-Al Prose does not ask for any passage to be retained or used for training. OpenAI's published API terms exclude training by default unless the customer opts in, but default abuse-monitoring logs may retain content for up to 30 days. Google's published paid Gemini API terms exclude product-improvement use of prompts and responses, but permit limited abuse-monitoring logging. We have not verified which upstream account agreements, logging controls or zero-data-retention approvals apply to these gateway requests. The authenticated gateway model catalog checked on 22 October 2026 labels the configured OpenAI Responses and Google chat-completions surfaces as zero retention; this is gateway-provided metadata, not an independently inspected provider agreement or an audit of gateway request logs. These general terms and catalog labels do not establish end-to-end zero retention for Lev-Al Prose; our non-retention commitment covers our application database and application-authored logs, not independently unverified gateway or provider systems.
• Stripe: payment and billing details for checkout and subscriptions. Stripe never receives passage text.
• Cloudflare Turnstile: a bot-protection check on the free passage. It never receives passage text.

Provider references: OpenAI API data controls; Gemini API retention guidance; Gemini logging policy. General terms reviewed on 22 October 2026; project-specific upstream configuration remains unverified.

Account records, usage counters and support conversations are protected by row-level access rules so that a signed-in person can reach only their own records. Administrators can access support conversations solely to answer them.

AI-detection scores are probabilistic and vary by detector, version, language and passage length. Lev-Al Prose improves naturalness and preserves authorship; it cannot guarantee a particular score or replace an institution's academic-integrity requirements.

You can request deletion of your account and its stored records through the in-platform support system. Because submitted passages are not stored in the application database, there is no application text archive to delete.