Security and deployment
Clear boundaries. Explicit access.
Meso-AI-Gateway runs in your Google Cloud environment. Your team controls access, approved model connections, and where audit records are stored.
The boundary model
Your infrastructure, with defined responsibilities
Meshesha has no default runtime access to your cluster or traffic. You decide which providers receive requests, where logs are stored, and whether to grant access for a support engagement.
Data flows you control
- ✓Prompts and model inputs
- ✓API keys and credentials
- ✓Audit logs and usage telemetry
- ✓Model traffic and responses
- ✓Identity and access records
What Meshesha delivers
- →Helm chart and container images
- →Configuration documentation
- →Deployment runbooks
- →Security boundary descriptions
- →Update paths and change logs
Your infrastructure controls
- ◆You operate the GKE cluster
- ◆You configure network policies
- ◆You manage workload identity
- ◆You define rate limits and guardrails
- ◆You choose log destinations and retention
Engineering practices
How we build and deliver
Access and assessment evidence
- ✓Access-control and logging patterns designed with NIST SP 800-171 requirements in mind
- ✓Documented configuration and operating responsibilities for security review
- ✓Least-privilege IAM guidance included in deployment documentation
- ✓Assessment evidence scoped to the deployed system and its configuration
Secure delivery practices
- ✓Dependency and container scanning in the delivery pipeline
- ✓Container hardening documentation
- ✓Documented update and rollback paths
- ✓Report security issues to contact@meshesha.solutions with the affected version and reproduction details
Model access and request checks
- ✓Access policy applied across the model endpoints you approve
- ✓PII detection and prompt-injection guardrail hooks configurable by the customer
- ✓Identity-aware usage records routed to customer-controlled systems
- ✓Audit logs structured for downstream compliance workflows
What to establish before deployment
Define the data types, model providers, retention rules, and controls required for your workload.
Document external provider connections and any support access. These choices determine where information can travel.
Treat certifications and authorizations as specific to the assessed system and organization; gateway deployment alone does not confer them.
Plan the deployment around your requirements
Review the model endpoints, data flows, access rules, and operating responsibilities before deciding how the gateway should fit your environment.