Nirad AI Fortress · AI Usage Control

Your people are sending company data to AI. See it. Stop it.

Aadhaar, PAN, GSTIN, source code, contracts, flowing into ChatGPT, Claude, and Gemini at machine scale, with no record of what left. AI Fortress brings every AI interaction under one policy. Inline at the edge. No endpoint agents.

Request a demo →

Agentless

On-prem & sovereign

Runs on the edge you already have

AI Fortress · Inline AI Inspection Live

Outbound prompt → ChatGPT

Support agent · draft customer reply"Here's the record: name Priya M, Aadhaar ████ ████ ████, PAN ████████. Draft a refund response."

Redacted inline. Aadhaar & PAN stripped before the prompt left the network. The agent still gets a usable answer.

Aadhaar · PANredacted

0 bytessensitive data sent

Loggedfull audit trail

The AI attack surface

The perimeter moved. No one owns what leaves the building.

Employees and autonomous agents push prompts, files, and API calls to public AI at machine scale. The data leaving is invisible: no visibility, no policy, no record.

Your own confidential files

AI Fortress indexes your private documents, so it blocks your specific contracts, designs, and IP when pasted into an AI tool, even in fragments.

The differentiator

Sensitive data exfiltration

Aadhaar, PAN, GSTIN, source code, secrets, and customer records, recognised and stopped before they leave.

Shadow AI

Unsanctioned AI tools used across the org with zero visibility. You can't secure what you can't see.

Prompt injection

Malicious inputs that manipulate AI systems, a growing risk as you deploy your own AI apps and agents.

What AI Fortress does

See it. Control it. Secure it.

Inline between your people and every AI service, enforcing policy in real time, in both directions.

01 · See

Every AI tool in use, surfaced.

A live picture of which AI services your org actually uses, sanctioned and shadow, with who's using them and what data is involved.

Shadow AI mapped, not guessed at
Per-user, per-tool, per-data-class breakdown

AI services detectedthis week

ChatGPT

142 users · engineering, support

Sanctioned

Claude

38 users · product

Sanctioned

Gemini

11 users · finance

Shadow

Perplexity

7 users · marketing

Shadow

02 · Control

Your policy, enforced in real time.

Granular rules in both directions: what data may leave for external AI, and who can see what from internal AI. Redacted or blocked before anything leaves.

Rules by data class, user, and tool
Real-time redaction, not after-the-fact reports

Active policyoutbound → external AI

aadhaar in any prompt Redact

source_code to public AI Block

api_key in any prompt Block

indexed_file match Block

general_text low risk Allow

03 · Secure

Internal AI that respects who's asking.

As you build internal assistants (HR bots, knowledge tools, RAG over private docs), AI Fortress enforces your existing access rights at the AI layer, so a response never reveals what the user isn't entitled to see.

Permissions enforced at the AI layer
Every interaction logged for audit

Internal HR assistantaccess enforced

Employee · no payroll access

What's the salary band for the senior backend role?

HR assistant

I can't share salary information. That's restricted to your role. Here's the job description and team structure instead.

Salary data withheld, enforced by AI Fortress, not left to chance

The product · one enforcement layer

One layer between your people and every AI service.

Every prompt, file, and API call inspected, decided, and enforced in real time, on the Nirad edge you already run.

Employee

Prompts & files

AI agent

API calls, machine scale

Nirad AI Control Plane

Inspect

Every prompt & call

PII scanner

Aadhaar · PAN · GSTIN

Injection guard

Prompt-attack defence

Policy engine

Rules-based

BlockRedactAllow

AI services

ChatGPT · Claude · Gemini

Ready Watching AI traffic in real time…

One vendor AI Fortress is a software capability on the Nirad SD-WAN + NGFW edge you already run. No new box, no new vendor. Every interaction logged, policy-linked, audit-ready.

Always on

AI Fortress works while you sleep.

Every AI interaction inspected, decided, and logged inline, with no one watching the console. Your audit trail writes itself.

inspected

blocked or redacted

100%

logged for audit

Live activity · last night

Illustrative activity. Live figures reflect your own environment.

Sovereign & compliant

Indian, on-prem, and DPDP-ready.

No foreign SaaS vendor governing your AI usage. Because AI Fortress already inspects every AI data flow, the compliance controls come for free.

On-prem by designDeploy inside your environment. Your document index never leaves. Nothing about your files is sent out to be inspected.

Made in India, MII / TTP-alignedBuilt and owned by the same Indian OEM behind your SD-WAN and NGFW.

DPDP alignment, for freeAudit trails and data-handling controls that support the DPDP Act, generated because AI Fortress already governs the flow.

MIIMake in India aligned

TTPAlignment

DPDP 2023Act-ready controls

On-premFull data residency

STQCIn progress

Common questions

What security teams ask first.

Does our data leave the organisation to be inspected?

No. AI Fortress runs on-prem, inline on the Nirad edge in your own environment. Inspection, redaction, and enforcement all happen inside your network, and the document index that recognises your confidential files stays on-prem. Nothing about your files is sent out.

Is there anything to install on endpoints?

No. AI Fortress is agentless. Because enforcement happens at the network edge you already run, every device and branch is covered by default, nothing to deploy or maintain on individual machines.

How does it sit inline without adding latency or breaking AI tools?

The Nirad SD-WAN and NGFW edge is already in the traffic path, so AI traffic flows through the same point of enforcement. AI Fortress inspects and applies policy in real time; where content is sensitive it's redacted or blocked, and the user still gets a usable response.

How is this different from generic data-loss prevention?

AI Fortress is built for the AI channel: outbound prompts to public AI and responses from internal AI. Beyond patterns like Aadhaar, PAN, and GSTIN, it indexes your own documents, so it recognises and protects your specific files even when only fragments are pasted in.