Security

Hosted workloads run on maintained, supported operating system versions.

Operating system and platform patches are applied on an ongoing basis.

Network-level firewalling restricts inbound access to hosted services.

Administrative and database access is restricted at the network layer.

Traffic to BMaiKR-operated applications is encrypted in transit via TLS.

Database access is limited to the application and authorized administrators.

Application access requires authenticated credentials.

Actions and data access are gated by the authenticated user’s permissions.

Where an application supports multiple roles, permissions follow a role-based model.

Accounts and services are scoped to the access they need to function.

User-supplied input is validated before processing.

APIs require authenticated, authorized requests.

Credentials and API keys are kept out of source code and application code repositories.

Access to production systems is reviewed periodically.

Hosted services are monitored for availability and errors.

Application and infrastructure events are logged for operational and security review.

Known vulnerabilities in dependencies and infrastructure are tracked and addressed.

Application dependencies are kept up to date as part of ongoing maintenance.

A documented process governs how security incidents are handled — see the Incident Response page.

Data moving between the customer, the application, and BMaiKR infrastructure is encrypted via TLS.

Applied where the underlying storage and hosting platform supports it.

Backups are access-restricted in the same way as production data.

Retention periods follow the applicable service configuration — see Customer Data Lifecycle.

Deletion follows the service’s configured procedure after the recovery window closes.

Client-delivered AI workflows are designed to supply only the context a specific task requires.

AI models are not given open, unrestricted access to a customer’s full database.

The AI provider used for a workload is a configuration decision, documented per project.

Context passed to a model is scoped and constructed by the application layer, not passed raw.

Retrieval-augmented workflows are designed to query only the knowledge base relevant to the task.

For applicable client-delivered platforms, access can be scoped to a specific customer’s data according to the agreed architecture.

Customer data and workflows can be kept separate at the application and storage layer where a platform’s architecture calls for it.

Integration credentials can be scoped to the customer and service they belong to.

Applied according to the deployment model agreed for a given platform.

Depends on the logging configuration deployed for a given platform.