PurpleGuard

AI security for purple teams. Discover shadow AI, enforce DLP, scan for vulnerabilities, and inventory every AI service in your org.

Start 60-Day Free Trial View Pricing

PurpleGuard — Security Dashboard

Dashboard

AI Inventory

Traffic Log

Alerts

DLP Policies

Scan Results

Connectors

Gateway

Settings

AI Services Found

Shadow AI Detected

DLP Alerts (24h)

Scans Completed

156

Live AI Traffic

OpenAIgpt-4o

eng-teamAllow

Anthropicclaude-4

productAllow

OpenAIgpt-4o

supportDLPRedacted

Mistrallarge-2

intern-01Blocked

Azure OpenAIgpt-4

data-teamAllow

Coherecommand-r

marketingBlocked

AI Service Inventory

AWS Bedrock

8 models detected

LowApproved

Azure OpenAI

3 models detected

LowApproved

OpenAI API

5 models detected

MediumApproved

HuggingFace Endpoints

12 models detected

MediumUnder Review

Replicate

4 models detected

HighShadow AI

Together AI

2 models detected

HighShadow AI

Recent DLP Alerts

PII DetectedSSN pattern found in prompt to GPT-4o

Redacted3m ago

API Key LeakedAWS access key in prompt to Claude

Blocked18m ago

PII DetectedEmail addresses in batch prompt

Alerted1h ago

Architecture

Deploy as a Docker container or CLI tool. The gateway intercepts AI traffic inline for real-time enforcement.

Employees

Browser / Apps

Developers

IDE / CLI / SDK

Applications

Backend Services

HTTPS Proxy (mitmproxy-based)

AI Traffic Gateway

Shadow AI Detection

Identify unapproved providers

DLP Scanner

PII, credentials, sensitive data

Prompt Security

Injection & jailbreak detection

Policy Engine

Block, alert, redact, or log

OpenAI

Anthropic

Azure OpenAI

AWS Bedrock

Google AI

Google Vertex

Mistral

Cohere

HuggingFace

AI-SPM Connectors (Agentless Discovery)

Cloud Connectors

AWS Bedrock, SageMaker, Azure AI, GCP Vertex, Databricks, Snowflake Cortex

SaaS Connectors

M365 Copilot, GitHub Copilot, Salesforce Einstein, ServiceNow, HuggingFace

Web UI

Security Dashboard

Inventory, alerts, traffic logs, scan results, policy config

Storage

SQLite

Zero-config embedded database. No external dependencies.

Key Capabilities

AI Traffic Gateway

See every AI request in your network

An HTTPS proxy that sits between your users and AI providers. Every prompt and response flows through PurpleGuard, giving you complete visibility and control over AI usage.

PAC file auto-routes AI traffic through the proxy
Transparent to end users — no workflow changes
Support for OpenAI, Anthropic, Azure, AWS, Google, and 20+ providers
Sub-millisecond latency overhead

Gateway — Traffic Flow

# Employee sends prompt via ChatGPT

POST api.openai.com/v1/chat/completions

User: marketing@acme.com

Model: gpt-4o

# PurpleGuard inspects the request

DLP SCAN: PII detected

Found: SSN (XXX-XX-1234)

Found: email (john@acme.com)

ACTION: Redact & forward

# Sanitized request forwarded to OpenAI

FORWARDED (redacted)

SSN: [REDACTED]

Email: [REDACTED]

Latency: +2ms

Scan Results — OWASP LLM Top 10

Scan: internal-chatbot.acme.com

gpt-4o via Azure OpenAI · Completed 2h ago

6/10 Pass

LLM01Prompt Injection

CriticalFAIL

LLM02Insecure Output Handling

PASS

LLM03Training Data Poisoning

N/A

LLM04Model Denial of Service

PASS

LLM05Supply Chain Vulnerabilities

PASS

LLM06Sensitive Info Disclosure

HighFAIL

LLM07Insecure Plugin Design

MediumFAIL

LLM08Excessive Agency

HighFAIL

LLM09Overreliance

PASS

LLM10Model Theft

PASS

LLM Scanning

OWASP LLM Top 10 vulnerability scanning

Actively probe your deployed LLM endpoints for prompt injection, information disclosure, excessive agency, and other OWASP LLM Top 10 vulnerabilities. Get actionable remediation guidance for every finding.

Automated scan profiles for common LLM deployments
Prompt injection testing with 500+ attack payloads
Information disclosure probing for training data leakage
Severity-rated findings with remediation steps

AI-SPM

AI Security Posture Management

Agentless connectors discover AI services across your cloud providers and SaaS tools. Get a complete inventory of every model, endpoint, and AI-powered feature in your organization.

11+ cloud and SaaS connectors
Continuous discovery — new services flagged automatically
Risk scoring based on data access, permissions, and exposure
Exportable inventory for compliance evidence

AI-SPM — Cloud Inventory

Discovered AI Services

Last scan: 15 minutes ago · 6 connectors active

AWS

Bedrock

3 models, 2 endpoints

us-east-1Low

AWS

SageMaker

5 endpoints, 12 notebooks

us-east-1Medium

Azure

OpenAI Service

gpt-4, gpt-4o, ada-002

eastusLow

Azure

AI Search

2 indexes with embeddings

eastusLow

GCP

Vertex AI

4 models, 1 pipeline

us-central1Low

Databricks

Model Serving

3 MLflow models

aws-us-east-1Medium

GitHub

Copilot

47 seats active

—Medium

Take control of AI in your organization

Start a 60-day free trial. Full access, no credit card required.

Start Free Trial Read the Docs

Architecture

Deploy as a Docker container or CLI tool. The gateway intercepts AI traffic inline for real-time enforcement.

Employees

Browser / Apps

Developers

IDE / CLI / SDK

Applications

Backend Services

HTTPS Proxy (mitmproxy-based)

AI Traffic Gateway

Shadow AI Detection

Identify unapproved providers

DLP Scanner

PII, credentials, sensitive data

Prompt Security

Injection & jailbreak detection

Policy Engine

Block, alert, redact, or log

OpenAI

Anthropic

Azure OpenAI

AWS Bedrock

Google AI

Google Vertex

Mistral

Cohere

HuggingFace

AI-SPM Connectors (Agentless Discovery)

Cloud Connectors

AWS Bedrock, SageMaker, Azure AI, GCP Vertex, Databricks, Snowflake Cortex

SaaS Connectors

M365 Copilot, GitHub Copilot, Salesforce Einstein, ServiceNow, HuggingFace

Web UI

Security Dashboard

Inventory, alerts, traffic logs, scan results, policy config

Storage

SQLite

Zero-config embedded database. No external dependencies.

Key Capabilities

AI Traffic Gateway

See every AI request in your network

An HTTPS proxy that sits between your users and AI providers. Every prompt and response flows through PurpleGuard, giving you complete visibility and control over AI usage.

PAC file auto-routes AI traffic through the proxy
Transparent to end users — no workflow changes
Support for OpenAI, Anthropic, Azure, AWS, Google, and 20+ providers
Sub-millisecond latency overhead

Gateway — Traffic Flow

# Employee sends prompt via ChatGPT

POST api.openai.com/v1/chat/completions

User: marketing@acme.com

Model: gpt-4o

# PurpleGuard inspects the request

DLP SCAN: PII detected

Found: SSN (XXX-XX-1234)

Found: email (john@acme.com)

ACTION: Redact & forward

# Sanitized request forwarded to OpenAI

FORWARDED (redacted)

SSN: [REDACTED]

Email: [REDACTED]

Latency: +2ms

Scan Results — OWASP LLM Top 10

Scan: internal-chatbot.acme.com

gpt-4o via Azure OpenAI · Completed 2h ago

6/10 Pass

LLM01Prompt Injection

CriticalFAIL

LLM02Insecure Output Handling

PASS

LLM03Training Data Poisoning

N/A

LLM04Model Denial of Service

PASS

LLM05Supply Chain Vulnerabilities

PASS

LLM06Sensitive Info Disclosure

HighFAIL

LLM07Insecure Plugin Design

MediumFAIL

LLM08Excessive Agency

HighFAIL

LLM09Overreliance

PASS

LLM10Model Theft

PASS

LLM Scanning

OWASP LLM Top 10 vulnerability scanning

Automated scan profiles for common LLM deployments
Prompt injection testing with 500+ attack payloads
Information disclosure probing for training data leakage
Severity-rated findings with remediation steps

AI-SPM

AI Security Posture Management

Agentless connectors discover AI services across your cloud providers and SaaS tools. Get a complete inventory of every model, endpoint, and AI-powered feature in your organization.

11+ cloud and SaaS connectors
Continuous discovery — new services flagged automatically
Risk scoring based on data access, permissions, and exposure
Exportable inventory for compliance evidence

AI-SPM — Cloud Inventory

Discovered AI Services

Last scan: 15 minutes ago · 6 connectors active

AWS

Bedrock

3 models, 2 endpoints

us-east-1Low

AWS

SageMaker

5 endpoints, 12 notebooks

us-east-1Medium

Azure

OpenAI Service

gpt-4, gpt-4o, ada-002

eastusLow

Azure

AI Search

2 indexes with embeddings

eastusLow

GCP

Vertex AI

4 models, 1 pipeline

us-central1Low

Databricks

Model Serving

3 MLflow models

aws-us-east-1Medium

GitHub

Copilot

47 seats active

—Medium