AIDR Overview
You can use AI detection and response (AIDR) to gain visibility into generative AI usage, detect threats, and enforce policies in enterprise environments.
Requirements
AIDR requires one or more subscriptions:
- AIDR for Workforce
- AIDR for Agents
These subscriptions can be combined with
NextGen SIEM .Default roles:
- AIDR Admin
- AIDR Viewer
Permissions for custom roles:
- AI Detection and Response
- Manage AIDR findings and agent collectors
- Manage AIDR findings and workforce collectors
- Read AIDR data from LogScale
- Read AIDR findings and agent collectors
- Read AIDR findings and workforce collectors
AIDR is available in these CrowdStrike clouds:
- US-1
- US-2
- EU-1
For more information,
Contact us .Capabilities
Visibility into AI activity
AIDR collectors capture AI interactions across browsers, applications, gateways, and cloud platforms. The collected telemetry includes prompts, responses, and metadata (user identities, device information, application context). You can correlate this data in dashboards and detailed logs to gain visibility into AI usage patterns across the organization, or view it in CrowdStrike Falcon Next-Gen SIEM for correlation with endpoint, network, and identity data.
LLM threat detection
AIDR detects the following risks in generative AI interactions, with optional policy enforcement:
- Prompt injection and jailbreak attempts - Adversarial prompts designed to manipulate AI behavior or bypass security controls
- Sensitive data exposure - PII, credentials, financial data, and confidential information in prompts and responses via built-in patterns, natural language processing, and custom definitions
- Malicious entities - Known malicious URLs, IP addresses, and domains in AI outputs using integrated threat intelligence
- Toxic and harmful content - Violent, abusive, or harmful content in AI inputs and outputs
- Language - Language detection with optional use of an allowlist or denylist
- Topic violations - Configurable content category restrictions
How AIDR works
Collectors
Collectors gather AI telemetry from different parts of the enterprise environment. Each collector type captures AI activity from a specific layer:
- Browser - Browser extension that captures user interactions with AI provider sites (ChatGPT, Claude, Gemini, and others) in managed browsers
- Application - SDK and API integration for instrumenting internal applications with inline AI security checks
- Agentic - MCP (Model Context Protocol) Proxy that captures AI traffic between MCP clients and servers
- Gateway - Network-layer proxy integration (Kong, LiteLLM, and others) that inspects AI traffic at API gateways
- Cloud - Cloud platform integration that ingests AI-related logs and events from supported platforms (AWS Bedrock)
- OpenTelemetry - Standardized telemetry instrumentation for collecting AI-related data from applications and services
You can configure collectors for different scenarios depending on deployment location. You can register collectors in the AIDR console and associate them with a security policy.
Policies
You can use AIDR policies to define what to detect and how to respond. Then, you can assign policies to collectors. Each policy contains two types of rules:
- Access rules - Attribute-based conditions that control how requests are processed based on metadata (user identity, device, application ID, and other attributes)
- Prompt rules - Content-based detectors that inspect prompts and responses for security risks
Each rule can be configured with an action:
- Log - Record the interaction without intervention (monitoring mode).
- Redact - Detect and replace sensitive content before submission or delivery (transform mode).
- Block - Prevent the request from reaching the AI model or user (block mode).
You can configure policies in the AIDR console. See Policy configuration for details.
Visibility and analysis
You can access logs of all collected telemetry, including:
- Original prompts and AI responses
- Request metadata (timestamp, user ID, device ID, application ID, collector ID, etc.)
- Detection results (identified risks, applied actions, redacted content)
You can access logs in the AIDR console through:
- Data flows and dashboards - Visualizations of AI usage patterns, relationships, detection trends, and policy enforcement outcomes
- Logs and Findings - Detailed logs of individual AI interactions with detection context and enforcement actions
- NextGen SIEM integration - AIDR saves logs to CrowdStrike Falcon Next-Gen SIEM for correlation with other security telemetry.
Use cases
Employee workforce monitoring
Security teams can deploy AIDR to monitor employee use of AI tools in managed enterprise environments. This use case applies to organizations where endpoints are managed (via MDM or enterprise tools) and network access is controlled (via proxies, secure web gateways, or zero trust solutions).
AIDR provides visibility into employee AI activity, detects sensitive data exposure and policy violations, and enforces content policies across managed browsers and gateways. Collector deployments may include Browser collectors on managed browsers, Gateway collectors at network proxies, and Cloud collectors for sanctioned cloud AI platforms.
AI application development
Application developers can integrate AIDR into AI-powered applications to implement inline security controls. This use case applies to internally developed AI systems, autonomous agents, customer-facing chatbots, and other AI applications that require threat detection and policy enforcement.
AIDR provides application-level instrumentation for logging AI interactions, detecting threats in prompts and responses, and enforcing policies before data reaches AI models or users. Collector deployments may include Application collectors (via SDKs or APIs), Agentic collectors (MCP proxy), Gateway collectors at API boundaries, Cloud collectors for cloud-based AI services, and OpenTelemetry instrumentation for standardized telemetry collection.