Security & Trust
Security and privacy are built into Hazel from the ground up. Review our security controls, compliance posture, and authorized subprocessors below.
All Systems Operational
29 of 29 controls passing
Zero Data Retention with AI model providers
Your data isn’t stored after it’s been processed. AI vendors delete it immediately after use and never use it to train their models.
No AI Model Training
Altruist and its service providers do not use your data to train, fine-tune, or improve AI models — ever.
You Control Your Data
Choose what gets recorded, how long it’s retained, and when it’s deleted — across video, audio, transcripts, and meeting summaries.
Meeting Transparency
Hazel must be admitted to join meetings and automatically notifies all participants via chat, so everyone has clear visibility.
Firm-Level Isolation
Hazel isolates your firm’s data at the account level. Your client data is never shared across firms or commingled with other accounts.
End-to-End Encryption
All data — chats, video, audio, transcripts, and summaries — is encrypted with TLS 1.2+ in transit and AES-256 at rest.
Security Controls
Access Control
Multi-Factor Authentication (MFA) Enforcement
Systems are configured to authenticate users via a unique user account, MFA, and minimum password requirements or SSH keys prior to granting system access.
Role-Based Access Control (RBAC)
Predefined security groups are used to assign role-based access privileges and segregate access to in-scope systems and data.
Privileged Access Restrictions
Administrative access privileges to in-scope systems are restricted to user accounts accessible only by authorized personnel.
Access Request & Manager Approval Workflow
Internal user access requests are documented in an automated ticketing system and require manager approval prior to access being granted.
Employee Termination Access Revocation
Termination checklists are completed and system access is revoked for all employees as a component of the termination process.
Quarterly Privileged Access Reviews
User access reviews, including privileged access, are performed by management on a quarterly basis to ensure access is restricted to authorized personnel.
Data Security
Encryption in Transit (TLS)
Web servers use TLS > 1.2 encryption for all web communication sessions. Unencrypted HTTP connections are rejected.
Endpoint Detection & Response (EDR)
An EDR application is installed on all entity-owned workstations to detect and respond to cyber threats in real time.
Encryption at Rest (AES-256)
Production data is stored encrypted using AES-256-bit encryption.
Data Handling & Encryption Policy
Documented policies and procedures guide personnel in the handling and encryption of stored data across all systems.
Infrastructure
Annual Third-Party Penetration Testing
Penetration testing is performed by a third-party vendor annually.
Monthly Vulnerability Assessments
Automated vulnerability assessments are performed on a monthly or more frequent basis.
Intrusion Detection System (IDS)
An IDS is configured to report network events.
Availability
Monthly Capacity Planning Reviews
Engineering personnel review infrastructure and system capacity monthly. Planning is conducted to mitigate the effects of infrastructure or system changes on availability.
Real-Time Infrastructure Monitoring
Logging and monitoring software collects data from system infrastructure components and endpoints to monitor performance, security vulnerabilities, and resource utilization, with alerts for unusual activity.
Automated Daily Backups
An automated backup system performs scheduled backups of all production data on a daily basis.
Business Continuity & Disaster Recovery (BC/DR) Plan
BC/DR plans are in place to guide personnel in procedures to protect against disruptions caused by unexpected events.
Annual BC/DR Plan Testing
A documented BC/DR plan is tested by in-scope functional areas on an annual basis to validate recovery procedures.
Annual Backup Restore Testing
Backup data restore tests are performed by IT personnel at least annually to verify backup integrity and recovery capability.
Change Management
Source Code Access Restrictions
Access to modify production system source code is restricted to user accounts accessible only by authorized personnel.
Pull Request Review Enforcement
Version control software is configured to enforce pull request review by personnel other than the change author before merging changes into production.
Peer-Reviewed & Approved Change Process
Application and infrastructure changes are authorized, peer reviewed, tested, and approved prior to implementation in production.
Segregated Development & Production Environments
Development and test environments are strictly segregated from production environments. Production access requires explicit authorization.
CI/CD Deployment Notifications
CI/CD software is configured to notify engineering personnel when changes are merged to production.
Vendor Management
Annual Vendor Compliance Monitoring
Management performs monitoring activities for all third-party vendors annually to ensure compliance with Hazel's security requirements.
Vendor Access Requirements Policy
Vendor management policies specify access requirements for all vendors and business partners with access to Hazel systems.
Incident Response
Post-Incident Postmortem Analysis
A postmortem analysis is performed for all security incidents, covering impact analysis, resolution, lessons learned, and tracked action items.
Incident Response Program
A documented IR program defines roles, containment, remediation, operational restoration, communication protocols, and lessons learned processes.
Documented Incident Escalation Procedures
Documented escalation procedures guide personnel in identifying and reporting security failures, incidents, concerns, and other complaints.
Subprocessors & Vendors
The following third-party service providers may process data on behalf of Hazel. Altruist and its service providers do not use your data to train, fine-tune, or improve AI models. When AI processing is used, it is done under Zero Data Retention (ZDR).
Zero Data Retention with AI model providers
Hazel maintains a zero data retention policy with AI vendors as a standard practice. Your data isn't stored after it's been processed — vendors delete it immediately after use and don't use it to train their models.
| Subprocessor | Description | Region | Data Retention |
|---|---|---|---|
| Anthropic | AI model provider. | United States | Zero data retention |
| Assembly AI | Convert speech to text. | United States | Zero data retention |
| AWS | Infrastructure provider. | United States West | Can be deleted upon request. |
| Azure | Document processing. | United States East | Zero data retention |
| Braintrust | AI observability and product quality monitoring. | United States | Zero data retention |
| OpenAI | AI model provider. | United States | Zero data retention |
| Recall | API to get transcripts and recordings from meetings. | United States East | Zero data retention |
- Zero data retention
Anthropic
AI model provider.
United States
- Zero data retention
Assembly AI
Convert speech to text.
United States
AWS
Infrastructure provider.
United States West
- Zero data retention
Azure
Document processing.
United States East
- Zero data retention
Braintrust
AI observability and product quality monitoring.
United States
- Zero data retention
OpenAI
AI model provider.
United States
- Zero data retention
Recall
API to get transcripts and recordings from meetings.
United States East
“Zero data retention” means AI subprocessors delete data immediately after first use, storing it only for the minimum time required to complete processing and never using it for model training. Because meeting audio and video files are voluminous and require resilient, fault-tolerant handling, our transcription partners may briefly store them solely to ensure reliable processing, after which they are promptly deleted.