Use Cases

See what Notis protects

Real scenarios. Real risks. One solution that works across every AI surface in your organization.

Provider-agnostic by design — works with OpenAI, Anthropic, Azure OpenAI, Google Gemini, and any OpenAI-compatible endpoint (Perplexity, Groq, DeepSeek, xAI, Together, OpenRouter, and more).

🏢

Enterprise-Wide Protection

No App Changes

Govern AI traffic across the whole organization — no SDK, no per-app changes

Risk: Only 5% of organizations report full visibility into their AI usage · Check Point, 2026 Cloud Security Report

Without Notis

✕IT can't see which apps, agents, or scripts are calling AI providers
✕Shadow AI spreads across engineering, support, and analytics — unmanaged
✕Each team rolls its own controls, or none at all
✕No central policy, no unified audit trail — every call is a blind spot

# IT security's view of AI egress:

$ dig api.anthropic.com
→ 160.79.104.10 (external, direct)

# Which app sent it? What was in the payload?
# No record. No policy. No way to know.

With Notis

✓Internal DNS routes traffic for your AI endpoints through Notis
✓Apps, developers, and agents covered transparently — no SDK, no per-app code changes
✓One CA, pushed through the MDM you already run, establishes trust org-wide
✓Centralized dashboard maps every AI data flow across the org

# 1. Route AI endpoints to the gateway (split-horizon DNS):
api.anthropic.com → notis-gateway.internal
api.openai.com    → notis-gateway.internal

# 2. Trust the Notis CA via existing device management:
$ mdm push notis-root-ca.pem   → all managed clients

# Every AI call from a managed client now flows
# through policy — no application changes required.

Deploy: Route your AI endpoints with DNS, then push the Notis CA through your existing MDM. Managed clients are covered automatically — no SDK, no per-app config.

The dashboard shows all traffic routed through Notis; pair with your egress monitoring to surface any direct-to-provider connections that bypass it.

🏦

Financial Services

GLBA / SOX

Account numbers, balances, and customer financial records flowing to AI providers

Risk: GLBA, SOX, PCI, and FINRA obligations apply to every AI call

Without Notis

✕Account numbers, balances, and SSNs sent to LLMs in plaintext
✕Advisory and trading prompts leak client identities and positions
✕No examiner-ready record of what customer data left the firm
✕Providers may retain regulated financial data for training

# Wealth-advisory assistant:
POST api.openai.com/v1/chat/completions

"Summarize this client's portfolio:
 Client: Margaret Chen
 SSN: 425-73-8291
 Account: 8847-201-93
 Balance: $2,480,019  Cash: $312,400"

With Notis

✓Account numbers and SSNs tokenized — format-preserving and reversible by you
✓The model sees realistic surrogates; real values never leave your boundary
✓Detokenize on the response so downstream systems get the true value back
✓Every regulated field, who touched it, and the action — in an exam-ready audit trail

# What the provider actually receives:
POST gateway.notis.cloud/v1/chat/completions

"Summarize this client's portfolio:
 Client: Margaret Chen
 SSN: 900-71-4412         ← reversible surrogate
 Account: 4021-887-55     ← reversible surrogate
 Balance: $2,480,019  Cash: $312,400"

Deploy: Tokenize regulated fields on the way out and detokenize on the way back — your systems keep working, the provider never sees a real account or SSN.

🪪

Consumer Data & Privacy

GDPR / CCPA

Names, emails, addresses, and SSNs flowing to AI across support, marketing, and product

Risk: GDPR penalties reach 4% of global annual revenue

Without Notis

✕Customer PII pasted into prompts across support, sales, and analytics
✕No record of which personal data went to which provider
✕Deletion and data-subject requests can't cover what you can't see
✕Cross-border AI calls move PII outside its allowed region

# Support copilot drafting a reply:
POST api.anthropic.com/v1/messages

"Draft a refund email to:
 Sarah Okafor
 sarah.okafor@gmail.com
 4821 Pinecrest Ave, Austin TX
 SSN on file: 425-73-8291"

With Notis

✓PII detected and acted on before it reaches the model
✓Per-field policy: redact emails, tokenize SSNs, monitor names
✓Every personal-data flow logged for DSAR and audit evidence
✓Consistent enforcement across every app and team — no SDK

# What the model actually receives:
POST gateway.notis.cloud/v1/messages

"Draft a refund email to:
 Sarah Okafor
 [EMAIL_1]
 [ADDRESS_1]
 SSN on file: [SSN_1]"

Deploy: Detect and act on PII at the gateway — redact, tokenize, or monitor per field — so personal data is governed the moment it leaves.

🏥

Healthcare AI Assistants

HIPAA

Patient records, clinical notes, and PHI flowing to AI providers

Risk: $1.5M+ per HIPAA violation

Without Notis

✕Patient SSN, MRN, DOB sent in plaintext to Claude/GPT
✕Medications, diagnosis codes, and provider notes exposed
✕No audit trail proving HIPAA data flow controls
✕Auditor asks 'what PHI left through AI?' — no answer

POST api.anthropic.com/v1/messages

"Summarize patient record:
 Patient: John Smith
 SSN: 425-73-8291
 MRN: 847-293-001
 Dx: Type 2 Diabetes (ICD-10: E11.9)
 Rx: Metformin 1000mg BID"

With Notis

✓SSN, MRN, DOB, ICD codes automatically redacted
✓Claude still produces useful clinical summary
✓Full OTEL audit trail with policy version and findings
✓Auditor gets: 'Here are our AI data flow controls and logs'

POST gateway.notis.cloud/v1/messages

"Summarize patient record:
 Patient: John Smith
 SSN: [SSN_REDACTED]
 MRN: [MRN_REDACTED]
 Dx: Type 2 Diabetes ([ICD_CODE REDACTED])
 Rx: Metformin 1000mg BID"

Deploy: Change one URL or add one DNS record. Try it live →

Try live demo →

🛡️

Government & Defense

CMMC / CUI

CUI, mission data, and export-controlled information reaching commercial AI providers

Risk: ITAR/EAR violations carry criminal penalties; CUI spillage is reportable

Without Notis

✕CUI and controlled technical data sent to commercial LLMs
✕ITAR/EAR-controlled details leave the authorization boundary
✕No boundary-level record of what crossed to which provider
✕Staff and contractor shadow AI evades policy entirely

# Analyst summarizing a controlled doc:
POST api.openai.com/v1/chat/completions

"Summarize this report:
 // CUI//SP-CTI
 Program: Sentinel UAV avionics
 Export control: ITAR USML Cat. XV
 Coordinates: 34.7°N, 117.9°W"

With Notis

✓Controlled markings and data detected and blocked at the boundary
✓Block, redact, or monitor per policy — controlled data never leaves raw
✓Boundary-level audit trail built for incident reporting and assessment
✓Runs as a dedicated, single-tenant gateway

# Notis blocks the egress:
POST gateway.notis.cloud/v1/chat/completions

→ 403 BLOCKED  policy: controlled-data-egress
  finding: CUI marking + ITAR reference
  logged: trace gov-4471, analyst id, full payload
  result: request never reached the provider

Deploy: Enforce controlled-data policy at the egress boundary — block or redact before anything reaches a commercial provider — with an audit trail built for assessment.

Air-gapped and on-prem deployment for classified environments is on the roadmap; today the gateway runs as a dedicated single-tenant instance.

🔄

Production AI Workloads

PCI-DSS

RAG pipelines, AI agents, and customer-facing features calling LLM APIs at scale

Risk: $4.88M average cost of a data breach

Without Notis

✕Customer support AI sends full customer records to GPT
✕Financial analysis agents include account numbers in prompts
✕RAG pipelines embed PII from vector databases into queries
✕Thousands of API calls per hour with zero inspection

# Customer support agent:
POST api.openai.com/v1/chat/completions

"Help customer with billing:
 Name: Jane Doe
 Email: jane@acme.com
 Card: 4532-0151-1283-0366
 Account: 9928374"

With Notis

✓Inline gateway inspects every request at sub-millisecond speed
✓SSE streaming preserved — no buffering, no latency impact
✓Per-workload policies and rate limiting
✓Every request logged with trace ID, findings, and actions

# What GPT actually receives:
POST gateway.notis.cloud/v1/chat/completions

"Help customer with billing:
 Name: Jane Doe
 Email: [EMAIL_REDACTED]
 Card: [CARD_REDACTED]
 Account: [REDACTED]"

Deploy: Point your provider base URL — OpenAI, Anthropic, Azure, or any OpenAI-compatible API — to Notis. Same SDK, same format, same response.

💻

Developer AI Tools

IP Protection

Claude Code and Codex sending source code and secrets to AI providers

Risk: API keys, credentials, and source code exposed

Without Notis

✕Developers paste database credentials into AI prompts
✕Source code with hardcoded API keys sent to external models
✕Internal architecture and trade secrets in code context
✕Security team has zero visibility into developer AI usage

# Developer asks Claude Code for help:

"Fix this config:
 DB_PASSWORD=super_secret_prod_123
 STRIPE_KEY=sk_live_abc123def456
 AWS_SECRET=AKIA..."

With Notis

✓Desktop app intercepts all AI tool traffic transparently
✓Credentials and API keys redacted before reaching provider
✓Developer experience unchanged — zero friction
✓Security team sees every request in the audit log

# What Claude Code actually receives:

"Fix this config:
 DB_PASSWORD=[ENV_SECRET REDACTED]
 STRIPE_KEY=[ENV_SECRET REDACTED]
 AWS_SECRET=[ENV_SECRET REDACTED]"

Deploy: Install desktop app. Click the shield. Every developer is protected.

Nothing leaves without Notis.

AI data protection for healthcare, finance, government, and enterprise. Deploy in minutes.

Try the Live Demo