Data Security and Privacy

Part 1 / Summary

Executive summary: Is our data safe?

The short answer: yes, your data stays yours, and only the people you approve can use it

Digital Brain is designed so that your confidential data stays under your control. It is protected in four ways: only people you approve can see it; the AI does not learn from it or remember employee conversations; it is not shared with other customers or used to train AI models; and it runs in a secured Microsoft Azure environment that is designed for enterprises.

Your questions, answered

Your question	Short answer	How it is protected
Who can see our data?	Only people you approve.	Three checks are required: a user must be provisioned with authorized credentials, must sign in to an approved workspace, and your workspace owner must separately assign which data that user may use. Login alone gives no access to company knowledge, and access can be changed or revoked at any time.
Will the AI learn from our data?	No, not by default.	The underlying AI model is not trained or fine-tuned on your documents. Your knowledge is only looked up when an authorized user asks a question. Any model training would be a separate customer project requiring your explicit approval.
Does the AI remember what our employees ask?	No.	Digital Brain keeps no persistent chat memory and does not learn from people’s conversations, so past conversations do not become a source for future answers. Instead, Digital Brain uses a human-defined enterprise context model to organize and connect business knowledge, enabling accurate, context-aware AI responses. This ensures that AI operates within a controlled, human-governed business context, rather than learning from unvalidated random conversations.
Could our data be shared with other companies or AI providers?	No.	Your documents and business information remain your property. For the Microsoft Azure AI services used, Microsoft states that prompts, responses and embeddings are not shared with other customers or used to train its models without permission.
Can our data be hacked or stolen?	Protected in layers, with settings agreed with you.	The solution runs on Microsoft Azure behind controlled sign-in and data assignment. Encryption, monitoring, backup, vulnerability management, support access and incident notification are managed in accordance with enterprise-grade security standards. The stored knowledge is treated as confidential even though it is broken into fragments.
Where is our data kept, and can we remove it?	USA, Europe, or a region of your choice (optional); export and deletion on request.	The hosting region and AI-processing location are fixed in your deployment plan. Export, deletion and retention terms are written in your service agreement.

Four design principles

Structured knowledge, not intact files

Source documents are broken into text chunks, vector embeddings and metadata instead of being kept as original files inside the AI knowledge database.

Stateless conversation design

Digital Brain does not use previous user conversations as persistent AI memory. Each request is resolved using controlled enterprise context.

Human-defined semantic context

A semantic engine connects enterprise information under a context model defined and governed by people, rather than relying on uncontrolled conversation learning.

Three layers of access control

A provisioned user signs in to a workspace; the workspace owner then assigns which company data is accessible to that user. Upon request, we can provide a dedicated server environment hosted on Microsoft Azure, along with a Super Admin application and role to centrally and securely manage access across the enterprise.

What this means for your organization

Digital Brain is designed for controlled access to enterprise knowledge, including business processes, transformation documents, application landscapes, project decisions and operating procedures. Company information is used to support authorized questions, not to build a persistent personal memory of employee conversations.

Important scope: Data transformation and stateless AI behavior reduce certain exposures, but they do not replace encryption, identity management, authorization checks, logging safeguards or customer-specific contractual controls. These are addressed through the deployment configuration in the Microsoft Azure environment.

Part 2 / Technical detail for IT and security teams

1. How customer documents become AI knowledge

Digital Brain does not keep the source document in its original file form inside the AI knowledge database. During ingestion, the source content is processed into smaller information elements and organized for semantic retrieval, or the relevant data is retrieved only when it is needed (triggered by the user—for example, Slack integration via API) for a specific AI conversation, without storing the data.

STEP 1Documents
Approved source material
STEP 2Chunk & vectorize
Text segments + numerical embeddings
STEP 3Tag & connect
Metadata + context relationships
STEP 4Knowledge database
Governed retrieval records

What the AI knowledge database contains

Information element	Description	Why it is needed
Text chunks	Extracted portions of source text; not the complete original file as one document.	Provide facts and passages that can be cited or summarized.
Vector embeddings	Numerical representations used for semantic similarity search.	Identify relevant information despite differing query wording.
Metadata / tags	Attributes such as business process, topic, source, team or other configured context labels.	Organize knowledge and support controlled filtering.
Semantic relationships	Connections among business objects and information within the human-defined context model (architected metadata of Enterprise Architecture).	Improve relevance across related processes, roles and systems.

A key security distinction

Processed does not mean anonymous or unreadable
The database contains a combination of text fragments, metadata, and numerical vectors. While vectorization is not encryption, the fragmented data is difficult to interpret outside the Digital Brain environment. However, individual text chunks may still contain confidential business information. Therefore, Digital Brain treats all processed knowledge as protected customer data, subject to appropriate access controls and data storage security measures.

2. Stateless AI with a human-governed context model

General-purpose AI chat services offer conversation history or memory-based personalization. Digital Brain takes a different approach for enterprise knowledge: it intentionally avoids persistent conversational memory and supplies relevant organizational context through its semantic engine instead.

Dimension	Digital Brain approach	Security benefit
Conversation memory	No persistent user conversation memory is used for future AI answers.	Unlike browser-based AI (e.g., ChatGPT), the AI does not learn from employee conversations.
Context source	Relevant information is retrieved from organized company knowledge and the context model.	Answers rely on the assigned and authorized knowledge base only.
Enterprise meaning	Human-defined semantic model connects roles, processes, systems, projects and related information.	Knowledge organization can be reviewed, configured, and managed by designated owners, rather than being autonomously managed by AI through an opaque, uncontrolled process. This ensures transparency, human oversight, and governance of enterprise knowledge.
Model training	No default foundation-model training or fine-tuning with customer content.	Customer knowledge is retrieved only for approved use and is not incorporated into the base model by default.
Governance	Workspace data assignment and context-model administration determine the available knowledge.	Access and knowledge relationships are governed by designated owners, outside the conversational memory layer.

What “stateless” means

Each question is handled without accumulating a permanent chat memory
Digital Brain does not depend on remembered past conversations to know the organization. Its semantic engine retrieves current, human-organized context for an authorized user. The corporate-defined context model provides enterprise context for accurate and grounded responses and works as a managed control layer of AI behavior.

3. User identity and workspace-owner data assignment

Digital Brain does not grant access to company knowledge merely because a user has an account. The workspace owner controls a separate data-assignment step that determines which enterprise knowledge is available to that user.

GATE 1User provisioned
Authorized ID + password + workspace ID
GATE 2Workspace login
Identity authenticated
GATE 3Owner assignment
Company knowledge made available
AT QUERYAuthorized retrieval
Only assigned knowledge enters AI context

Right context for each user

Controlled access by design: reference model data, customer model data, source data, applications and views, workspaces, user authorization levels, and roles.

Access behavior by scenario

User scenario	Workspace access	Company knowledge available in AI chat
No valid credentials	Denied	None.
Valid user ID / password but no workspace access	Denied	None.
Authorized workspace user with no owner-assigned data	Allowed	No assigned enterprise knowledge for retrieval; the chat may still support general questions depending on settings.
Authorized workspace user with owner-assigned data	Allowed	Only the enterprise knowledge allocated to that user, subject to configured permission enforcement.
Owner changes or revokes data assignment	Account access may remain	Changes to a user’s role automatically revoke access to all previously assigned knowledge data.

Roles and responsibilities

Role	Responsibility
Workspace Owner	Approves and assigns available data to authorized users; updates or revokes assignments.
Workspace User	Signs in using provisioned credentials and asks questions within the scope of assigned data. Individual users can also upload personal data, which remains accessible only to them.
Super Administrator (Available with selected deployment options)	Supports system configuration, customized applications and views, and operational management, subject to applicable access controls, support policies, and customer governance requirements.

Security advantage: Authentication answers “Who is this user?” Workspace-owner assignment separately answers “Which enterprise data may this user use?” All three gates (provisioning, workspace login and owner assignment) are necessary for the governed company-knowledge retrieval.

4. Data protection and AI provider boundaries

Digital Brain supports an Azure-based architecture and can be integrated with customer environments such as SAP BTP/ERP. The exact storage locations, AI endpoints / MCP and operational responsibilities are documented for the selected customer deployment.

Security topic	Digital Brain position / deployment consideration
Customer data ownership	Customer-provided documents, data, and business information remain the customer’s content. Digital Brain’s proprietary context model and software are separate intellectual property.
AI model data use	The default RAG architecture does not train the underlying foundation model on customer data. Any customer-specific model training is a separate activity requiring explicit approval.
Cloud and model provider	Azure AI Foundry is a referenced backend in existing Digital Brain deployments.
Data storage	Processed text chunks, embeddings, and metadata are stored in the configured knowledge environment. Source file handling, backups, logs, and related storage are managed separately. Original documents and data are not retained in Digital Brain unless requested or agreed upon by the customer.
Geographic processing	The default deployment regions are the United States. Upon request, Digital Brain can be deployed in other regions supported by Microsoft Azure. Deployment outside Azure-supported regions can be discussed as a customized deployment option.
Access control	Provisioned user accounts, workspace permissions, and workspace-owner data assignments determine which enterprise knowledge, and context models each user can access.
Retention and deletion	Customer data retention, export, and secure deletion services are provided. By default, accounts and associated data are archived for one year. Upon request, customer data can be exported in an agreed file format or securely deleted.

5. Questions frequently raised by enterprise security teams

Customer question	Digital Brain answer
Are our documents stored as complete files in the AI database?	No. The AI knowledge database organizes processed text chunks, vector embeddings and metadata rather than maintaining original documents as intact files. Operational source copies, if any, are handled separately.
Is vectorized knowledge protected information?	Yes. Numerical embeddings and associated text or metadata remain customer information and require appropriate access and storage protection.
Does the AI remember employees’ previous chats?	No persistent conversation memory is used for future Digital Brain answers. Operational records, if any, are governed separately from AI conversational memory.
Does our information train the base AI model?	No. Optional model training requires an explicit project decision and approval.
Can anyone with a login see all company data?	No. Workspace entry does not automatically grant knowledge access. The workspace owner must separately assign the available enterprise data.
What if no data is assigned to a user?	No company knowledge is available for that user’s retrieval-based answers. The system may still respond to general queries depending on configuration.
Can Digital Brain provide source-grounded answers?	Digital Brain uses retrieval and source/context associations to support answers. Like all generative AI, outputs should be reviewed when used for material decisions.
Can we export or delete our information?	Yes. Upon request, customer data can be exported in an agreed file format or securely deleted. By default, accounts and associated data are archived for one year; exact retention and deletion terms are defined in the service agreement.

6. Customer security summary

Our approach to protecting enterprise knowledge

Digital Brain is designed to transform source documents and data into organized AI knowledge rather than placing intact documents in its AI knowledge database. Its stateless conversational architecture avoids persistent user-chat memory. Its human-defined semantic model determines the organizational context used for retrieval, while a three-layer access model requires a provisioned user account, an authenticated workspace login, and explicit workspace-owner data assignment before company knowledge becomes available to the user.

About this report

This overview describes Digital Brain’s stated product design and the security principles described in prior customer documentation, updated with the current product information supplied by Rock15. It is intended for customer architecture and security review; it is not a certificate, penetration-test attestation, or guarantee that no risk exists. Detailed technical and contractual assurances are subject to the agreed deployment and service terms.

References

Microsoft Foundry — Data, privacy and security: https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/openai/data-privacy
Microsoft Azure — Shared responsibility in the cloud: https://learn.microsoft.com/en-us/azure/security/fundamentals/shared-responsibility/
Microsoft Foundry — Model deployment types and data residency: https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/deployment-types

Product basis: Enterprise_Questionnaire_AI_251028.docx (previous customer questionnaire) and updated Digital Brain design descriptions supplied October 2026.