Before deploying AI-generated code, apply a specific security checklist (SQL injection, exposed secrets, input validation) — don't assume it's secure just because "it works".