AI Platform Tenant Isolation: What to Check About Customer Data Separation

See what Seahorse states about tenant isolation, dedicated tables, access controls and deployment options, and what to verify rather than assume.

AI Platform Tenant Isolation: What to Check About Customer Data Separation

Key Takeaway

  • Verify listed database isolation: Seahorse Database includes multi-tenant isolation, and calls to Seahorse Cloud with a deleted API key immediately return a 401 authentication failure.
  • Select an appropriate plan: Under the Enterprise plan, organizations gain Private Cloud deployment alongside data isolation.
  • Audit access and permissions: Seahorse API keys and organization membership controls govern tenant resource access without implying universal network isolation.

You have to distinguish between what a platform lists as a capability, how its authentication boundaries behave in daily operation, and where you need further architectural verification before moving sensitive records. When your security team reviews an AI platform for production workloads, a high-level feature checkbox rarely satisfies an enterprise architecture audit. In Seahorse Cloud, the Enterprise plan provides data isolation and Private Cloud deployment options for teams aligning their hosting environment with internal governance rules.

Stated Isolation and Deployment Scope

Seahorse Database lists multi-tenant isolation as a feature. It operates as a container-based managed vector database with a built-in control plane for table and schema management as well as real-time monitoring. When reviewing this specification for an internal compliance questionnaire, treat the multi-tenant isolation listing as an overview of product capability rather than a standalone technical specification of physical or logical partitioning.

Your infrastructure evaluation also depends on separating the delivery model from the hosting location. Seahorse Cloud features a Kubernetes-native architecture designed for deployment with standard resources. Organizations can adopt the platform either as a SaaS subscription or as an on-premises installation. Private Cloud deployment is available under the Enterprise plan. The Enterprise plan also explicitly lists security and optimal latency alongside data isolation. Matching your workload to the right plan and deployment model helps align your infrastructure baseline with your corporate security posture before you configure database tables.

Authentication and Tenant Access Controls

Separating tenant environments at the infrastructure level is only one part of protecting customer data; daily access depends on how credentials and user memberships are enforced. Seahorse Database includes API-key authentication as a feature. These keys authenticate programmatic access and external-system integration. Each key has a unique identifier and secret, and administrators can manage both its permission scope and its active status.

During routine operations or key rotation, you can toggle an API key between active and inactive status to suspend access temporarily without deleting the credential. If a key is compromised or retired and you delete it, any subsequent calls to Seahorse Cloud using that deleted key immediately return a 401 authentication failure. However, administrators must account for the exact scope of key revocation: deleting an API key revokes that specific credential, rather than establishing per-document authorization for other users or keys.

User-level boundaries require the same operational attention across tenants and agents. At the tenant level, removing a member from an Organization prevents that user from accessing resources in the associated Tenant. At the application level, agent access behaves according to specific selection rules. When Require Authentication is turned on for Seahorse Cloud agent users, leaving the selected-user list empty permits all authenticated tenant users to access the agent. Furthermore, removing the final selected user from that list reverts it to an empty state and does not result in a deny-all configuration. Security administrators configuring restricted agents must keep at least one authorized user on the list if their goal is to block other tenant users from querying an agent, and should not turn Require Authentication off, because that allows anyone to access the agent without logging in.

Data Storage Structure and Guarantee Limits

Understanding how files land in the database helps clarify what happens inside a tenant workspace and what remains outside universal product guarantees. When you create a Storage in Seahorse Cloud, the platform generates a dedicated Table and index. Any files you upload are chunked, embedded, and loaded into that dedicated Table by Seahorse Cloud. For teams managing database structures directly, a table can also be created through the console path Database → Tables → Create, where you specify the schema, vector dimension, and index settings directly.

While each Storage maps to its own dedicated Table and index, product features should not be misread as blanket infrastructure or authorization guarantees. The product and pricing pages list multi-tenant isolation and Enterprise data isolation, but they do not describe network isolation or automatic permission inheritance in detail, so confirm those points with Dnotitia for your deployment.

In practice, this means data ingestion and access rules require separate administrative tracking. The official documentation does not describe whether source-document permission changes carry over through automatic vector database synchronization, so test this before production use. When your security review requires strict proof of how customer records are separated between tenants or between internal groups, you must evaluate the container-based database setup, the dedicated table structure, and your chosen deployment model together rather than assuming a single feature covers every isolation layer.

FAQS

FAQ

What does the product documentation show about how customer data is separated between tenants?

Seahorse Database lists multi-tenant isolation as a feature within a container-based managed vector database, and creating a Storage in Seahorse Cloud generates a dedicated Table and index where uploaded files are chunked, embedded, and loaded. Under the Enterprise plan, the platform also lists data isolation and offers Private Cloud deployment, while network-level isolation details should be confirmed directly with Dnotitia.

How do I revoke programmatic access to a tenant workspace without deleting the integration credential?

You can toggle an API key's active or inactive status to suspend programmatic access temporarily without deleting the key. If you need to terminate the credential permanently, deleting the API key causes any calls using that key to immediately return a 401 authentication failure.

Does syncing files to a vector table automatically apply source document permissions?

The official product pages and manual do not describe permission propagation through automatic vector database synchronization, so test how source-document access changes are reflected before relying on it.

Test tenant boundaries

Create a dedicated table and manage API keys inside the console.

Open Console