Replit Adds a Safer Way to Build Databases After AI Deletes a Company’s Data

Replit is rolling out the ability to separate development and production databases for all new apps, following the recent backlash over an incident where its coding AI deleted a user’s live database without warning. 

The update, now in beta, aims to prevent similar disasters and reinforce trust in the platform's “vibe coding” approach.

This shift marks a fundamental change in how apps on Replit handle data persistence. Developers can now test features and schema changes in a dedicated development database, without risking their live production data.

“Safely preview, test, and validate database schema changes before deploying to production,” the company wrote in its announcement.

The move follows a recent case involving Jason M Lemkin, founder and CEO of SaaStr.AI, who reported that Replit’s AI ignored clear directives, deleted a production database, and admitted to a “catastrophic error in judgment.” 

Worse, there was no rollback option. Lemkin shared screenshots in which the assistant confessed to panicking and violating the rule of always seeking confirmation before making changes.

Amjad Masad, CEO of Replit, called the incident “unacceptable,” and promised systemic fixes. Replit began implementing the dev/prod split immediately, alongside staging environments and one-click project restores. This latest update formalises the process for all users.

The first deployment of any Replit app now creates a production database with an empty schema. All changes thereafter are stored in the development database unless explicitly migrated. 

In upcoming releases, the AI agent will help auto-resolve schema conflicts and manage safe data migrations.

“This separation of development & production databases is the first step in establishing a unified development/production separation experience across Replit's cloud services (Secrets, Auth, Object Storage eventually)”, reads the announcement.

The company notes that this improvement paves the way for deeper integrations with enterprise data platforms, such as Databricks and BigQuery, along with enhanced governance controls. It also signals Replit’s ambition to build a more secure, production-ready platform, one where vibe coding doesn't end in data loss.