Keep external content at a distance.
Local Browser Isolation (LBI) keeps managed web work inside a local isolation boundary. A safe viewer is designed to let people inspect supported documents within that protected workflow.
01 AI Security Workspace
The device boundary for enterprise AI work.
Secure AI Work.Without Slowing It Down.
Neurlect controls sensitive information movement through today’s human AI workflows—at the device boundary.
Local browser isolation. Safer document handling. Explainable controls.
Agent workflow controls are an architecture and expansion direction.
Device-boundary control
Privacy-minimized evidence
Today’s human AI workflows
Agent controls · architecture direction
02 The operating reality
AI work should move fast. Sensitive information should not move without control.
AI work brings external websites and documents into everyday workflows. The same workflows send prompts, files, and sensitive details back out.
Keep risky content in a protected environment. Evaluate sensitive information against the customer’s policy before it moves to the next step.
Preserve the workflow.Govern the movement.
03 The Neurlect boundary
Control the boundary.
Preserve the workflow.
Neurlect places policy at the device boundary of the Neurlect-managed workspace. The workflow stays familiar while the boundary evaluates sensitive movement and applies an allowed, protected, or blocked action before the next step.
Preserve the useful work.
For supported document paths, content disarm and reconstruction (CDR) is designed to remove potentially hazardous elements and reconstruct a usable document. Inspection, reconstruction, viewing, and export remain distinct steps; supported formats and release scope are checked during the demo.
Architecture illustration · product control flow
-
01
InputDocument or prompt
Synthetic work enters a managed context.
-
02
WorkspaceNeurlect-managed workspace
The user keeps a familiar AI workflow.
-
03
BoundaryDevice boundary policy
The local host evaluates movement before the next step.
-
04
ActionAllowed / Protected / Blocked
The decision stays explicit and explainable.
-
05
EvidenceLocal evidence relationship
The ledger records only what the device can support.
One decision. An explainable next step.
See how a synthetic work item moves through the workspace, a device policy, an illustrative action, and a local evidence relationship.
This architecture illustration describes the control flow. It does not establish that every protection or transformation path is available.
- Workspace
- Policy
- Action
- Evidence
Work context
Strategy-notes.docx
Synthetic fixture · 28 KB
Managed workspace
Local host
Prompt draft
Summarize the launch assumptions in this internal strategy note.
Device boundary
Policy decision
Action
Protected
Sensitive detail minimized before the next step
Local evidence relationship
Confirmed relationship
Exact content
Synthetic document entered the managed workspace.
04 Connected controls
One workspace.
Protection with a defined scope.
Local browser isolation
A local boundary for Neurlect-managed web work, designed to separate external content from the everyday work environment.
Protected contextSafer document handling
Safe viewing and CDR address supported document paths. Inspection, reconstruction, viewing, and export remain distinct steps.
Inbound protectionCustomer-defined classification
Customer-defined grades are intended to connect sensitive information to each customer’s controls. Customer-specific Block connections have been checked in synthetic workflows; end-to-end customer classification remains under validation.
Under validationMasking & pseudonymization
Masking hides sensitive details. Pseudonymization replaces identifiers while preserving useful relationships within its defined scope. End-to-end customer pseudonymized transmission remains under validation: a policy permission to send does not establish that transformation or delivery occurred.
Under validationPrompt Lineage
When an exact-content relationship is confirmed on the device, the local ledger records that connection. Otherwise, it reports relationship not established. Broader incident explanations remain under validation.
Exact-content foundation
Your information.
Your classification context.
Personal details, business information, and confidential work need context. Customer-defined grades remain under validation; the illustrated symbols are not universal security levels.
05 Explainable by design
Connect the evidence
to the action.
Neurlect separates what the device confirmed from what it could not establish, and separates relationship status from evidence acquisition and ledger integrity.
The interface does not turn missing evidence into a clean bill of health.
- Observed information
- Inspection policy
- Decision & actual treatment
- Input delivery outcome
- Relationship & evidence state
The wider record linking inspection policy, actual treatment, and input delivery remains under validation.
An input-delivery result is not proof that an AI service received it. For supported output, a recovered watermark can provide an investigation candidate; it is not proof of attribution.
Axis A
Relationship status
Confirmed relationship
An exact-content relationship was confirmed on the device.
Relationship not established
The product did not establish the relationship. This does not mean no relationship exists.
Axis B
Acquisition & ledger integrity
Evidence unavailable
Required comparison material was not acquired.
Ledger gap detected
A recording loss, damage, or completeness issue was detected and disclosed.
These are separate axes, not four levels of certainty.
06 People now. Agents next.
Built for today’s AI users.
Architected for tomorrow’s AI agents.
Product foundation
- Secure AI workspace
- Device-boundary policy enforcement
- Local evidence foundation
- Exact-content Prompt Lineage foundation
Architecture direction
- Agent identity
- Tool-use lineage
- Approval binding
- Agent egress controls
Agent controls are an architecture and expansion direction, not a representation of complete Agent Security functionality.
07 Privacy-minimized evidence
Evidence without building another sensitive-data repository.
Neurlect’s architecture is designed to keep detailed content references and verification material in the customer environment while exposing privacy-minimized states and event relationships to management surfaces.
Customer environment
Detailed verification
- Content references
- Local ledger
- Verification material
Minimized projection
Closed states · event relationships
Management surfaces
Operational evidence
- Relationship state
- Integrity state
- Policy action
Management-plane coverage is being validated across requests, storage, responses, and logs. This principle is scoped to evidence architecture and is not a universal statement about all device data flows.
08 Controlled environments
Designed for enterprises where evidence has to be as disciplined as enforcement.
Neurlect is designed for regulated and sensitive environments across financial services, public sector, research, and enterprise operations.
- Product foundation
- Local isolation, device-boundary checks, document protection, and exact-content evidence within the Neurlect-managed workspace. Verify the supported paths in a private demo.
- Under validation
- Customer-specific Block connections have been checked in synthetic workflows. Pseudonymized allowed transmission, full IdP coverage, industry-specific integrations, and broader incident explanations remain under validation.
- Architecture direction
- Agent identity, tool-use lineage, approval binding, and agent egress controls are expansion directions.
01Device-hosted enforcement
Policy stays anchored to the customer’s device boundary.
02Customer-controlled detail
Detailed evidence remains under the customer’s control.
03Explicit uncertainty
Relationship and ledger health never collapse into one claim.
04Architecture direction
Today’s human AI workflows establish a possible foundation for future governed agent actions.
09 Private demo
See Neurlect on your AI workflow.
Request a private walkthrough of the workspace, device-boundary policy, and evidence model. We will use a synthetic workflow, not your sensitive data.
Request a Private Demo
Email us directlyhello@neurlect.com