Security overview
How Avra protects your data — the security program, isolation, encryption, access, audit, who does what, and how to report a vulnerability.
Avra runs a security program aligned with the NIST Cybersecurity Framework (CSF) 2.0, and is pursuing ISO 27001:2022 certification. Avra isn’t certified yet.
This page explains how Avra protects your data at each layer: how it’s kept apart from other customers, how it’s checked and encrypted, who can reach it, and what you control yourself. Each section links to the page with the details.
Avra's Trust Center
Avra’s controls, compliance status, and subprocessors.
Discover
At a glance
| Area | What Avra does | Read more |
|---|---|---|
| Isolation | Isolates your data by tenant and workspace, in the environment you choose | Your data stays yours |
| Encryption | Encrypts customer data in transit and at rest | Encryption |
| Data at ingestion | Validates every delivery against your agreed data contract before a model can use it | Data at ingestion |
| Access | Grants access to its systems and to customer data based on roles and need-to-know | Access and audit |
| Audit | Records account activity and every prediction in the dashboard, exportable as CSV | Audit and usage logs |
| Development | Defines its infrastructure as code, in version control, and scans its code and dependencies for known vulnerabilities | Development |
| Privacy | Processes data in compliance with LGPD | Data privacy & compliance |
| Reporting | Takes vulnerability reports privately at security@avra.ai |
Vulnerability disclosure policy |
Your data stays yours
We don’t share information between customers.
Your data is isolated by tenant and workspace, in the environment you choose, and nothing about your business trains another customer’s model. When Avra runs in your environment, what returns to Avra is execution metadata, such as each command’s status and result; operational telemetry labeled with your cluster name, which covers logs from Avra’s pods, the metrics and traces of Avra’s components, and resource metrics for the collector’s node and the pods on it; and the deployment agent’s release status and logs. Your data, the models adapted to it, and their outputs stay in your environment. See Data privacy & compliance, and the data layer’s security details.
Three boundaries keep your data where it belongs. The tenant boundary separates you from every other customer, and Avra doesn’t share information between customers. Inside your tenant, your organization holds your members and their roles, which decide what each person can do in the dashboard, and the models available to you. And each workspace keeps its own files, batches, API keys, and logs apart from your other workspaces.
How a delivery is protected
Each layer below has its own section on this page. Together, they cover your data from the moment it’s delivered to the predictions made from it.
Encryption
Customer data is encrypted in transit and at rest. When Avra runs in your environment, adaptation and fine-tuning artifacts are written to your storage, encrypted.
Data at ingestion
Every delivery is validated against your agreed data contract before a model can use it. A delivery that fails is rejected, not published. See Data ingestion.
Access and audit
Access to Avra’s systems and to customer data is role-based and granted on a need-to-know basis. In the dashboard, each member’s role decides what they can do in your organization. See Roles & permissions.
The dashboard records account activity — API keys, members, workspace and organization settings — under Observability → Audit, and every prediction under Observability → Usage, both exportable as CSV. See Audit and usage logs.
To control access on your side, see Team management for members and API keys for the keys your systems use.
Development
Avra’s infrastructure is defined as code, in version control, and Avra scans its code and dependencies for known vulnerabilities.
Privacy
Avra’s processing is LGPD-compliant. Data protection terms, retention, and deletion are set out in your agreement. See Data privacy & compliance.
Who does what
This table sums up what you do and what Avra does. Each row comes from a section of this page or the page it links to.
| You | Avra | |
|---|---|---|
| Your data | Remains your exclusive property | Acts as a data processor on your behalf |
| Data contract | Agree on it with Avra | Validates every delivery against it before a model can use it |
| Isolation and encryption | Isolates your data by tenant and workspace, and encrypts it in transit and at rest | |
| Access to your organization | Give each member a role, and remove access people no longer need | Applies each member’s role in the dashboard and through the API |
| Avra’s own access | Grants access to its systems and to customer data based on roles and need-to-know | |
| API keys | Rotate or delete any key that should stop working | Ties each key to one workspace |
| Records | Review the audit and usage logs, and export them as CSV | Records account activity and every prediction |
| Running in your environment | Host Avra’s data layer next to your data, and Avra RFM in your AWS account | Keeps the control plane, and receives execution metadata and operational telemetry labeled with your cluster name |
| Retention and deletion | Set in your agreement | Follows your agreement |
| Vulnerabilities | Report them privately to security@avra.ai |
Scans its code and dependencies for known vulnerabilities |
Status and reporting
Live status and incident history are at status.avra.ai. Report vulnerabilities privately to security@avra.ai. See the Vulnerability disclosure policy.