Security Model

Traditional enterprise access decisions often answer whether a person, application, device, or service is authorized to access a resource.

AI introduces a different context.

A legitimate employee may use an AI application or agent with different trust, processing, destination, or data-handling characteristics.

D[AI]NGEN therefore treats AI as a separate access context.

The Windows user keeps their normal access rights. D[AI]NGEN is not designed to remove the employee’s ordinary permissions.

Each AI application receives its own independent and more restricted access scope.

The AI application may reach its authorized Workspace and explicitly approved resources. Everything else in the user’s data scope remains inaccessible to that AI application by default.

Approval for one AI application does not automatically transfer to another AI application.

The public D[AI]NGEN model is:

Enterprise Data → D[AI]NGEN → Context / Policy → Decision → AI

Relevant conceptual context may include:

Discovery & Classification surfaces AI-relevant files, folders, categories, and sensitivity signals locally, then places what needs attention into an administrator review flow.

Discovery does not grant access by itself. An administrator or authorized decision-maker reviews and approves the resulting access scope.

Employees can use their ordinary files and request additional AI access, but cannot independently broaden the AI application’s scope.

A management decision and an enforced endpoint state are not treated as the same thing.

The product model requires the endpoint application of the intended access state before that state is considered effective.

Local filesystem resolution and raw content analysis are designed to remain on the endpoint where supported.

Central administration expresses intent and receives the safe metadata and product evidence required for review.

The product model is designed so a newly requested AI share does not silently override an explicitly protected scope.

Conflicting access is intended to fail closed rather than expanding AI access implicitly.

This is a public conceptual security model.

It describes the intended security boundary and decision flow.

It does not imply that every input, policy engine, enforcement mechanism, or decision outcome shown is currently implemented as a verified shipping feature.