Binding through Key Terms Part 3.1. It states the measures in place for Customer's instance of the Cloud Service.
| # | Measure | In place |
|---|---|---|
| 1 | Encryption at rest | Cloud-provider encryption of database storage, object storage and backups; application-level AES-256-GCM encryption of personal-data fields annotated in the data schema; an encryption failure stops the write rather than storing plaintext |
| 2 | Encryption in transit; network | All external traffic over HTTPS/TLS. The application reaches the database over a private network, and the database has no public IP address |
| 3 | Dedicated instance and tenant separation | Each customer runs on its own application service and its own database, with its own encryption keys and its own web address. Within that instance, database row-level security and tenant checks in the application remain in place as an additional safeguard. |
| 4 | Access control | Role-based permissions on every API operation; multi-factor authentication required for all users, including administrators, in customer tenants; sessions expire after 8 hours of inactivity and at an absolute limit of about 24 hours, validated server-side on each request; separation-of-duties controls on compensation changes. Provider personnel use multi-factor authentication for their company accounts and administrative access |
| 5 | Audit logging | Security-relevant and data-changing operations are logged and integrity-signed; database triggers block changes to audit content and deletion before the retention period ends |
| 6 | Application protections | Protection against common web attacks, including request-rate limits, strict input validation, cross-site request forgery protection and security headers |
| 7 | Change management | Automated security testing before changes reach production; staged deployment through development and staging before production, with rollback |
| 8 | AI safeguards | All AI calls pass through one gateway; automated name detection replaces many personal names before content reaches an AI provider (detection is not guaranteed); each user can turn AI processing off, and Customer's administrators can turn it off for the whole organization in the Cloud Service's settings; AI Output is advisory and labeled in the web application |
| 9 | Backups | A database backup is taken before every schema-changing deployment, and deployment halts if it fails. Automated daily backups are retained for 30 days, with point-in-time recovery. No multi-region replication is promised |
| 10 | Monitoring | Error tracking with personal-data scrubbing before transmission; external uptime checks of each customer instance; security and availability alerts paged to on-call personnel through PagerDuty; health checks |
| 11 | Incident response | Written incident-response and breach-notification procedures; customer notice per Key Terms Part 5 |
| 12 | Physical security | Provided by Google Cloud at its data centers; Provider operates no data-bearing facilities |
| 13 | Personnel | All personnel with access to Customer Content and Customer Personal Data are bound by written confidentiality obligations before access is granted |
| 14 | Vendors | Subprocessors bound by written data-protection terms; list and notice per Key Terms Part 5 |
Assurance status. Provider has not completed an independent security audit or penetration test and is preparing for a SOC 2 examination. Written security documentation and questionnaire answers are available under confidentiality.