Skip to main content

Vendor Security Assessment

Updated today

Prepared by: Blaze Analytics vGmbH

This document provides pre-filled answers to common vendor security assessment questions. If your organization uses a specific security questionnaire or vendor risk platform (SIG, CAIQ, HECVAT, etc.), contact [email protected] and we will complete it for you.

Registration

LU35935057

Address

23 Boulevard Friedrich Wilhelm Raiffeisen, 2411 Luxembourg

Product

BlazeSQL — AI-powered SQL data analyst

Enterprise contact

2. Service Description

BlazeSQL connects to customers' SQL databases and answers data questions in natural language. It generates SQL queries, executes them, and returns results as tables, charts, and dashboards.

Key architectural principle: BlazeSQL never imports or copies customer databases in full. To generate queries, it accesses schema metadata (table names, column names, data types). With offline mode enabled (default on Desktop), only schema metadata leaves your device. When offline mode is disabled, query results are also sent for deeper AI analysis. Actual data rows are queried on demand.

Encryption: All data stored on BlazeSQL servers is encrypted with AES-256 at rest and TLS 1.2+ in transit. Zero Data Retention (ZDR) is enabled on all AI model calls.

Deployment options:

  • Desktop app — query results stay local on the user's device by default (offline mode enabled)

  • Web app — results stored encrypted on BlazeSQL's GCP infrastructure

  • Database Connection API — all query processing stays in the customer's own infrastructure


3. Data Handling and Classification

What data does BlazeSQL access?

  • Database schema metadata (table/column names, data types) — always

  • Database credentials — all deployment models, stored encrypted. Exceptions: SQL Server Windows Authentication uses device credentials; Entra Authentication is token-based, no stored credentials

  • Query results — web app stores server-side; desktop app stores locally (offline mode, on by default); DB Connection API stores in customer infrastructure

  • Chat messages — stored encrypted on BlazeSQL servers

  • Unique values for categorical columns (column samples) — optional, stored encrypted

Is customer data used for AI model training?

No. BlazeSQL does not train models on customer data. Zero Data Retention (ZDR) is enabled on all AI model calls via GCP Vertex AI, meaning prompts and responses are not stored by Google and are not used for model training.

How is customer data used?

Solely to provide the BlazeSQL service: answering questions, generating SQL, displaying results, enabling dashboards and collaboration.

Data classification:

BlazeSQL processes data across a range of sensitivity levels depending on the customer's database contents. This is why BlazeSQL offers multiple deployment models — customers with highly sensitive data (PHI, student records, financial NPI) can use the Desktop app or Database Connection API to keep data rows entirely within their own environment. Schema metadata and chat messages are classified as confidential and are encrypted at rest and in transit.


4. Encryption and Key Management

Control

Implementation

Encryption at rest

AES-256 across all GCP data centers

Encryption in transit

TLS (HTTPS) for all client-server communication

Database encryption (Firestore)

AES-256 at rest, TLS in transit, custom security rules

Key management

Google-managed encryption keys by default

Hardware security modules

Cloud HSM available for enterprise (FIPS 140-2 Level 3)


5. Access Control and Authentication

User access model:

  • Database access is controlled by the admin who added the database

  • Only explicitly invited users can access a database

  • Permission levels: read-only or read-write (set by admin)

  • All users require unique accounts

Authentication methods:

  • Standard: email and password

  • SSO: SAML 2.0 and OpenID Connect

  • Multi-Factor Authentication (MFA): Two-factor authentication (2FA) is available natively. MFA is also enforced through your identity provider when using SSO (SAML/OpenID Connect).

  • Service principal / enterprise identity provider integration available (e.g., Microsoft Entra ID)

Internal access (BlazeSQL employees):

  • Production data access is restricted to personnel required for service operation and support

  • All production access is logged and reviewed regularly

  • Access is revoked promptly upon departure or role change


6. Network Security

  • BlazeSQL servers run in a firewalled, non-public subnet on GCP

  • Web app uses a static IP address for customer firewall whitelisting

  • All communication encrypted via TLS

  • GCP's built-in DDoS protection and network security controls apply


7. Vulnerability Management and Security Testing

Penetration testing:

Application-layer penetration testing is conducted by a qualified third-party firm. Results are available to enterprise customers under NDA. BlazeSQL's infrastructure inherits GCP's continuous security testing program.

Dependency scanning:

GitHub Dependabot scans for known vulnerabilities in third-party dependencies.

Bug reporting:

Security issues can be reported to [email protected].


8. Software Development Lifecycle (SDLC)

BlazeSQL follows secure development practices:

  • Code review: All code changes require peer review via GitHub pull requests before merge.

  • Dependency management: GitHub Dependabot monitors third-party dependencies for known vulnerabilities.

  • Deployment: Deployments are managed through controlled release processes. Production changes require code review approval before merge.


9. Physical Security

BlazeSQL has no physical infrastructure. All computing, storage, and networking runs on Google Cloud Platform data centers. GCP data centers are certified under ISO 27001, SOC 2, and other standards, and include physical security controls such as biometric access, 24/7 monitoring, and environmental protections. See Google's data center security documentation.


10. Incident Response

BlazeSQL follows a structured six-step incident response process:

  1. Identification: GCP security monitoring and real-time alerting flag anomalous activity

  2. Assessment: Security team evaluates severity, scope, and potential impact

  3. Containment: Affected components isolated immediately to limit exposure

  4. Eradication: Root cause identified and eliminated

  5. Recovery: Services restored using GCP backup and recovery infrastructure

  6. Post-Incident Review: Root cause analysis, effectiveness review, prevention planning

Notification: Affected customers are notified promptly in accordance with applicable regulations (e.g., GDPR requires notification within 72 hours of becoming aware of a personal data breach).


11. Business Continuity and Disaster Recovery

  • BlazeSQL runs on Google Cloud Platform, which provides built-in redundancy across availability zones

  • Data is stored in GCP Firestore with automatic replication

  • Backups are taken daily and retained for 30 days

  • Recovery Point Objective (RPO): 24 hours

  • Recovery Time Objective (RTO): 7 days

  • GCP's backup and recovery infrastructure supports service restoration

  • For enterprise deployments, custom backup and DR configurations are available


12. Compliance and Certifications

GCP Infrastructure Certifications (held by Google Cloud Platform, not BlazeSQL)

Certification

Held By

Scope

SOC 1, SOC 2, SOC 3

Google Cloud Platform

Security, availability, and confidentiality controls

ISO 27001, ISO 27017, ISO 27018

Google Cloud Platform

Information security and cloud security management

GDPR

Google Cloud Platform

EU data protection regulation

CCPA

Google Cloud Platform

California Consumer Privacy Act

HIPAA

Google Cloud Platform

Health Insurance Portability and Accountability Act

FIPS 140-2 Level 3

Google Cloud Platform

Cryptographic key management (via Cloud HSM)

BlazeSQL's compliance posture:

  • GDPR compliant (data subject rights, data export within 14 days, deletion controls, subprocessor transparency)

  • CCPA compliant

  • HIPAA-ready (BAA with Google in place; customer BAAs available for enterprise contracts)

  • Zero Data Retention verified on all AI model calls


13. Third-Party / Subprocessor Management

Subprocessor

Service

Data Processed

Location

Google Cloud Platform

Cloud infrastructure, compute, storage

All service data

EU / US (configurable)

Google Vertex AI

AI model inference

Chat prompts (ZDR enabled — not retained)

EU / US

Google Cloud Firestore

Application database

Account data, schema, chats, query results

Per GCP region

Crisp

Customer support live chat

Name, email, support conversations

EU (Netherlands / Germany)

Stripe

Payment processing

Payment details, billing email, transaction data

US

BlazeSQL will notify customers of material changes to subprocessors.


14. Data Retention and Disposal

  • Active accounts: Data retained until user deletes it or closes account

  • Self-service deletion: Users can delete any stored data — including chats, queries, dashboards, results, database connections, credentials, and schema metadata — at any time

  • Bulk deletion: Organization-wide deletion available via support

  • Account closure: All associated data deleted, subject to any legally required retention periods

  • Source databases: Never modified or affected by BlazeSQL deletion actions

  • Data export: Machine-readable export available within 14 days upon request

  • Backups: Retained for up to 30 days following deletion, then permanently removed

  • Server logs: Retained for a minimum of 1 year. Available for audit purposes or deleted upon request.


15. Employee Security

  • Production system access is restricted to personnel required for service operation

  • Unique accounts required; shared credentials are not permitted

  • All production access is logged

  • Employees with access to customer data are bound by confidentiality and non-disclosure agreements

  • Access is revoked promptly upon departure or role change, including deprovisioning of all accounts and credentials

  • All team members receive security awareness onboarding. Access to production systems requires completion of security training.


16. Additional Information

Static IP for whitelisting: Available for web app database connections

Database Connection API: Query processing stays entirely within customer infrastructure — maximum data isolation option

Dedicated GCP region: Available for enterprise data residency requirements


Contact

For security assessments, questionnaire completion, or additional documentation:


© Blaze Analytics vGmbH (LU35935057), 23 Boulevard Friedrich Wilhelm Raiffeisen, 2411 Luxembourg

Did this answer your question?