AICT security analyzer AI agent
This AI agent generates context for a detected AI security event by aggregating asset, prompt, cross-threat, and correlation data. It then computes a confidence score from detection strength, event risk, and authorization signals.
Workflow
For all threats, the agent deterministically executes outcomes in a single finalization step: creating a security incident for critical severity findings with no open duplicate, saving a triage record capturing the full verdict and rationale, and triggering an AI insight for platform-wide visibility after the triage record is committed.
- Retrieve the detected AI security event record and gather its full context, including the related asset, prompt, cross-threat indicators, and recent threat correlation.
- If the event context indicates an error, finalize the analysis immediately with the error details and stop.
- Calculate a confidence score using detection strength, event type risk, prompt authorization signals, and any cross-threat or correlation overrides.
- Finalize the analysis by determining severity and creating a security incident with no open duplicate. Save a triage record with the full rationale, and trigger an AI insight for platform-wide visibility.
- If finalization fails, present the error and its details to the user.
|
Field |
Description |
|---|---|
|
Allow third party to access this AI agent |
When enabled, third-party AI agents can use this agent. This value is off (false) by default. This setting is defined in the AI Agent configs [sn_aia_agent_config] table on the External discoverable field. |
|
Allow AI specialists to access this AI agent |
When enabled, AI specialists can use this agent. This value is off (false) by default. When set to true, more configuration options for tools become available so that an AI specialist can map inputs and response templates to tool outputs. This setting is defined in the AI Agent configs [sn_aia_agent_config] table on the Specialist enabled field. |
|
Manage long-term memory |
When enabled, all previous user interactions are used as context for the LLM. This value is off (false) by default. This setting is defined by the sn_aia.ltm.enable_long_term_memory system property. For more information, see ServiceNow Otto AI agents reference. |
|
Tools |
Assess Threat Finalize Analysis |
|
Allowed user roles The specific user roles that can access this AI agent. |
sn_ai_governance.ai_steward |
|
Data access roles The specific user identity roles that determine which data the AI agent can access and what actions it can take. |
sn_ai_governance.ai_steward |
|
Triggers |
Optional. None defined by default. An admin can specify triggers if desired. For more information, see Add a trigger to an AI agent. |
|
Channels |
Configure an assistant for Virtual Agent or ServiceNow Otto panel using Assistant Designer. |
|
Used in agentic workflows |
Not applicable. |
AICT Security Analyzer agent and Post-runtime settings
Post-runtime probabilistic metrics require that the AICT Security Analyzer agent is turned on. Both the AI agent and probabilistic metrics are turned off by default, and the sampling rate for the metrics is set to 1%. Turn them on to get additional monitoring and analysis of your AI systems.
Metrics turned off by default
The following metrics are turned off by default so that no AI usage is consumed until you choose to turn them on. For more information about configuring these metrics, see Configure post-runtime security metrics.
Warning: AI judge model classifications are probabilistic in nature and may be incomplete or incorrect. They don't constitute professional advice and shouldn't be relied on as the sole basis for assessing risk.
- Agent goal deviation
- Correctness detection
- Data integrity incident detection
- Excessive agency
- Improper input and output handling
- Prompt injection
- Sensitive data disclosure
- System prompt leakage
--- title: AICT security analyzer AI agent description: This AI agent generates context for a detected AI security event by aggregating asset, prompt, cross-threat, and correlation data. It then computes a confidence score from detection strength, event risk, and authorization signals.Post-runtime probabilistic metrics require that the AICT Security Analyzer agent is turned on. Both the AI agent and probabilistic metrics are turned off by default, and the sampling rate for the metrics is set to 1%. Turn them on to get additional monitoring and analysis of your AI systems. locale: en-US canonical_url: https://www.servicenow.com/docs/r/intelligent-experiences/plat-ai-aict-security-analyzer-agent-ai-agent.html release: brazil topic_type: reference last_updated: "2026-09-10" reading_time_minutes: 3 keywords: [ServiceNow Otto, AI Agents, generative AI, agentic AI] breadcrumb: [ServiceNow AI Platform AI agents, ServiceNow AI Platform, AI agents library, AI agents and agentic workflows, Enable AI Experiences] --- # AICT security analyzer AI agent This AI agent generates context for a detected AI security event by aggregating asset, prompt, cross-threat, and correlation data. It then computes a confidence score from detection strength, event risk, and authorization signals. ## Workflow For all threats, the agent deterministically executes outcomes in a single finalization step: creating a security incident for critical severity findings with no open duplicate, saving a triage record capturing the full verdict and rationale, and triggering an AI insight for platform-wide visibility after the triage record is committed. 1. Retrieve the detected AI security event record and gather its full context, including the related asset, prompt, cross-threat indicators, and recent threat correlation. 2. If the event context indicates an error, finalize the analysis immediately with the error details and stop. 3. Calculate a confidence score using detection strength, event type risk, prompt authorization signals, and any cross-threat or correlation overrides. 4. Finalize the analysis by determining severity and creating a security incident with no open duplicate. Save a triage record with the full rationale, and trigger an AI insight for platform-wide visibility. 5. If finalization fails, present the error and its details to the user. <table><thead><tr><th> Field </th><th> Description </th></tr></thead><tbody><tr><td> Allow third party to access this AI agent </td><td> When enabled, third-party AI agents can use this agent. This value is off \(false\) by default. This setting is defined in the AI Agent configs \[sn\_aia\_agent\_config\] table on the External discoverable field. </td></tr><tr><td> Allow AI specialists to access this AI agent </td><td> When enabled, AI specialists can use this agent. This value is off \(false\) by default. When set to true, more configuration options for tools become available so that an AI specialist can map inputs and response templates to tool outputs. This setting is defined in the AI Agent configs \[sn\_aia\_agent\_config\] table on the Specialist enabled field. </td></tr><tr><td> Manage long-term memory </td><td> When enabled, all previous user interactions are used as context for the LLM. This value is off \(false\) by default. This setting is defined by the **sn\_aia.ltm.enable\_long\_term\_memory** system property. For more information, see [ServiceNow Otto AI agents reference](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/intelligent-experiences/na-aia-reference.md). </td></tr><tr><td> Tools </td><td> - **Scripts** Assess Threat Finalize Analysis </td></tr><tr><td> Allowed user roles The specific user roles that can access this AI agent. </td><td> sn\_ai\_governance.ai\_steward </td></tr><tr><td> Data access roles The specific user identity roles that determine which data the AI agent can access and what actions it can take. </td><td> sn\_ai\_governance.ai\_steward </td></tr><tr><td> Triggers </td><td> Optional. None defined by default. An admin can specify triggers if desired. For more information, see [Add a trigger to an AI agent](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/intelligent-experiences/add-trigger-aia.md). </td></tr><tr><td> Channels </td><td> Configure an assistant for Virtual Agent or ServiceNow Otto panel using [Assistant Designer](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/conversational-interfaces/configure-now-assist-va.md). </td></tr><tr><td> Used in agentic workflows </td><td> Not applicable. </td></tr></tbody> </table>**Parent Topic:**[ServiceNow AI Platform AI agents](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/intelligent-experiences/platform-ai-agents-overview.md) ## AICT Security Analyzer agent and Post-runtime settings Post-runtime probabilistic metrics require that the AICT Security Analyzer agent is turned on. Both the AI agent and probabilistic metrics are turned off by default, and the sampling rate for the metrics is set to 1%. Turn them on to get additional monitoring and analysis of your AI systems. ### Metrics turned off by default The following metrics are turned off by default so that no AI usage is consumed until you choose to turn them on. For more information about configuring these metrics, see [Configure post-runtime security metrics](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/brazil/markdown/intelligent-experiences/gov-sec-configure-event-metrics.md). **Warning:** AI judge model classifications are probabilistic in nature and may be incomplete or incorrect. They don't constitute professional advice and shouldn't be relied on as the sole basis for assessing risk. - Agent goal deviation - Correctness detection - Data integrity incident detection - Excessive agency - Improper input and output handling - Prompt injection - Sensitive data disclosure - System prompt leakage
| title | AICT security analyzer AI agent |
| description | This AI agent generates context for a detected AI security event by aggregating asset, prompt, cross-threat, and correlation data. It then computes a confidence score from detection strength, event risk, and authorization signals.Post-runtime probabilistic metrics require that the AICT Security Analyzer agent is turned on. Both the AI agent and probabilistic metrics are turned off by default, and the sampling rate for the metrics is set to 1%. Turn them on to get additional monitoring and analysis of your AI systems. |
| locale | en-US |
| canonical_url | https://www.servicenow.com/docs/r/intelligent-experiences/plat-ai-aict-security-analyzer-agent-ai-agent.html |
| release | brazil |
| topic_type | reference |
| last_updated | 2026-09-10 |
| reading_time_minutes | 3 |
| keywords | ServiceNow OttoAI Agentsgenerative AIagentic AI |
| breadcrumb | ServiceNow AI Platform AI agentsServiceNow AI PlatformAI agents libraryAI agents and agentic workflowsEnable AI Experiences |