The vulnerability stems from improper sanitization in admin API response handlers. Both versions failed to properly exclude hidden attributes marked in content-type configurations. The v3 sanitizeEntity function and v4 permissions-manager sanitize function were both patched in the referenced PRs (#13189 and #13185) to implement attribute filtering. The exploit demonstrates attackers could manipulate filter parameters (e.g., changing 'email_containss' to 'password_containss') to bypass intended data restrictions. The SQL Injection CWE mapping likely refers to column name injection rather than traditional SQLi, as attackers manipulated attribute names in filter operations to access hidden fields.