Corey security and privacy

Corey is not trained on customer data.

It uses Azure OpenAI base models provided through Microsoft Azure AI Foundry. At runtime, Corey uses CoreView-developed prompts, orchestration logic, and platform skills to answer user requests.

To generate responses, Corey may use tenant data available in the CoreView platform, but only at runtime, within the authenticated user’s session, and according to the permissions and policies applied to that user. This data is used to answer the request, not to train or fine-tune the model.

Corey runs within a controlled Azure infrastructure using Azure AI Foundry. Data is processed only within the execution context of your authenticated session and organization.

CoreView uses Azure OpenAI Zero Data Retention (ZDR) for all Corey AI deployments. This means prompts and responses are not stored by the AI provider and are used only for the time required to process the request.

A separate concept is Data Zone Standard (DZS), which refers to the geographic area where inference processing takes place. At the moment, DZS is available only in certain Azure regions, currently the EU and the US. For customers who use Corey in regions where Azure does not yet provide DZS, CoreView may use Azure deployments that process inference in another supported region, according to the service configuration and contractual terms agreed with the customer.

Within Azure AI Foundry documentation, this capability is referred to as Global Standard deployment. With this deployment model, the model can still be configured within a customer's Azure Data Zone while Azure routes inference requests across its global infrastructure to the most appropriate region for processing, while remaining under Microsoft’s enterprise security and compliance controls.

Microsoft documentation reference:
Microsoft Learn – Deployment types for Microsoft Foundry Models

CoreView enforces strict tenant and session-level isolation. AI interactions occur within the context of your organization and authenticated user session, so requests and responses are always scoped to your environment.

Corey uses the same authentication and security layer used across the CoreView platform, which means the platform’s built-in security and access controls apply to AI interactions as well. Because of this architecture, tenant segregation and access controls prevent another customer or tenant from accessing your environment or its data through prompts.

CoreView uses Azure OpenAI Zero Data Retention (ZDR), which means prompts and responses are not stored by the AI provider and are used only for the time required to process the request.

On the CoreView side, interactions and queries may be logged and stored within the CoreView platform for security, auditing, governance, and service functionality purposes, including making conversations available in the product. This allows CoreView tenant administrators to review AI interactions if needed.

At this time, limited access may be available to authorized CoreView personnel for system monitoring and troubleshooting purposes. This access is restricted to CoreView only and does not extend to AI providers such as Azure AI Foundry or Azure OpenAI.

Access is strictly controlled and used only to support system reliability, service improvement, and issue resolution during the beta period.

CoreView follows enterprise-grade security practices and compliance standards aligned with the broader CoreView platform security program. The infrastructure runs on Microsoft Azure and follows Microsoft’s security model, while CoreView maintains industry-recognized security certifications and practices.

More information is available here:
https://www.coreview.com/security

CoreView applies rigorous validation processes and controls to maintain response quality and align results with the authenticated user context, available platform data, and the CoreView permission model.

We apply multiple layers of protection:

1. Azure AI Content Safety protections built into the Azure OpenAI platform.
   https://azure.microsoft.com/en-us/products/ai-services/ai-content-safety
2. Custom LLM guardrails implemented within Corey’s prompts and orchestration layer.
3. Azure Prompt Shield and jailbreak detection capabilities are used as part of the protection model.
   https://learn.microsoft.com/en-us/azure/ai-services/content-safety/concepts/jailbreak-detection

Data access through Corey is limited to the authenticated user’s organizational context and the CoreView permission model. The AI layer does not bypass platform permissions, which means Corey can access and return only information that the user is already authorized to access within CoreView.

Additionally, we use tool-level permission controls to allow tenant administrators to define which Corey tools specific users can access, providing more granular governance over the actions users can perform.

Yes. Communications with Azure services are encrypted in transit using TLS, and Azure infrastructure provides encryption and security controls. Zero Data Retention and Azure’s service architecture mean prompts are processed without persistent storage by the AI provider.

Corey can execute certain management actions within the CoreView platform, but these actions are never executed silently; they always require explicit user confirmation.

In many cases, supported actions have corresponding reverse actions within the platform. If such an action exists, it can be used to revert the change, again requiring user confirmation.