Redacted case study. How a privilege-escalation bug in AI-generated code was found and fixed before launch.
A founder built a B2B SaaS with Cursor in a few weeks and was days from launch. The app handled customer data but had never been security reviewed.
A five-day AI-Code Audit Sprint: a manual review of routes, controllers, models, auth and config, backed by static analysis and dependency scanning.
The user profile update bound request input directly to the model:
Because $fillable included the role column, any user could set role=admin on their own account. Four criticals in total, including a raw-query injection and secrets committed to the repo.
Every finding shipped with a fix. The team remediated in two days, we retested, and the app launched clean.