Data Is Isolated by Design
SAP AI services are built as tenant-aware environments, meaning each customer operates in a separate and isolated space.
In SAP AI Core, resources such as executions, deployments, configurations, and artifacts belong to a specific tenant or resource group and cannot be shared across tenants or resource groups. Workloads run in sandboxed environments and cannot access resources from other tenants.
This separation is enforced at the infrastructure and runtime level, ensuring that:
- your workloads run independently
- your data is processed within your own environment
- other customers cannot access your resources
This is the foundation of data protection in SAP Business AI.
Data Is Not Shared Across Customers
By default, your data is not shared with other customers and is not used for cross-customer improvements.
For example, SAP states for AI services such as Document AI that:
- data remains within your tenant
- optional feedback or learning features are generally not enabled by default
This reflects a consistent approach across SAP Business AI: customer data remains isolated within the tenant and is governed according to the selected service configuration and customer decisions.
Data Processing Depends on the AI Scenario
Not all AI works in the same way, and this affects how data is processed.
- In many embedded AI scenarios, processing happens within the application and is part of the workflow. Depending on the use case, AI may run in the application or through SAP-managed services such as SAP AI Core.
- In side-by-side scenarios, data is processed in SAP-managed environments on SAP Business Technology Platform (BTP), for example using services such as SAP AI Core, where AI workloads are executed in a controlled and isolated environment.
- In generative AI scenarios, requests are routed through SAP-managed services (such as the Generative AI Hub), where access to AI models is governed and controlled. Models are accessed through SAP-managed infrastructure, ensuring that interactions follow defined security and governance policies.
In these generative scenarios, SAP applications do not directly connect to external providers. Instead, requests are routed through controlled SAP services, where governance mechanisms such as access control, policy enforcement, and filtering can be applied before and after model interaction.
The selection and preparation of data included in AI requests (for example, limiting input to relevant data or masking sensitive information) is typically handled at the application level, based on the specific use case and business requirements.
This approach ensures that even when external models are used, data processing remains controlled, and exposure is minimized according to defined policies and configurations.
Clear Boundaries for Data Access
SAP defines clear boundaries for how data is accessed and processed.
Within a tenant, access to resources is controlled through authentication and authorization mechanisms. Across tenants, strict isolation ensures that no cross-customer access is possible.
In addition, SAP security measures protect data:
- from unauthorized access through authentication and authorization
- from unintended exposure through access controls and monitoring
- from misuse by ensuring data is processed only for the intended purpose
These controls help ensure that data is handled securely throughout its lifecycle.
Your Role as a Customer
In many enterprise AI scenarios, SAP provides the infrastructure and processing environment, while customers remain responsible for deciding how their business data and personal data are used, managed, and governed.
This means customers remain responsible for:
- deciding what data is used
- ensuring compliance with data protection regulations
- managing storage and deletion in connected systems
Secure infrastructure supports compliance, but governance remains a customer responsibility.
Lesson Summary
Your data is not shared with other customers by default. It is isolated within your tenant and protected through strict separation mechanisms.
At the same time, different AI scenarios involve different data flows, from fully contained processing within applications to controlled interaction with external models. Understanding these differences helps you make informed decisions about how your data is used in AI.