Threat Dashboard
Overview
The Guardrails dashboard provides a centralised view of security threats detected across AI agents, prompts, and tool invocations. The Guardrails dashboard focuses on runtime agentic security risks, including prompt injection, unauthorised tool access, sensitive data leakage, and malicious agent behaviour.
Guardrail States and Lifecycle
The main table lists all detected agentic security threats. Each row represents a consolidated security finding derived from one or more prompts, tool calls, or agent executions.
Guardrails are organized into the following operational states:
Active Guardrails that require investigation or remediation.
Under Review Guardrails currently being analyzed or validated.
Ignored Guardrails marked as non-actionable for the environment.
This lifecycle enables structured triage while preserving full execution context.
Guardrail Attributes Displayed
Each guardrail entry includes the following attributes:
Severity
Represents the risk level of the agentic guardrail based on exploitability and potential security impact.
Guardrail Name
Identifies the detected agentic security issue, such as prompt injection, excessive sensitive data exposure, or unauthorized tool access.
Detection Type
Indicates the detection category used to identify the violation, such as single-prompt analysis or tool invocation analysis.
Number of Endpoints
Indicates the count of agent-accessible endpoints or tools affected by the violation.
Domains
Lists the domains or agent platforms where the guardrail activity was observed.
Compliance Mapping
Shows the governance and compliance frameworks impacted by the detected agent behavior.
Discovered
Displays the timestamp when Akto first detected the violation.
Compliance-Based Guardrail Filtering
Compliance Framework Selector
The Guardrails dashboard allows filtering agentic threats by compliance framework to support governance and audit workflows.
Supported frameworks include:
CIS Controls
CMMC
CSA CCM
FedRAMP
FISMA
GDPR
HIPAA
ISO 27001
NIST 800-53
NIST 800-171
SOC 2
OWASP
This filter helps you focus on agent behaviours that directly impact regulatory and internal security requirements.
Navigating to Guardrail Activity Details
From the Guardrails dashboard, you can navigate to detailed guardrail activity views for request- and prompt-level investigation.
Select a guardrail from the Guardrails dashboard opens the Guardrail Activity Detailed View, which provides execution-level visibility into how the guardrail was detected.
From the detailed view, you can analyse:
Prompts associated with the guardrail
Tool invocations triggered by the agent
Sensitive data accessed or exposed
Guardrails violated during execution
Repeated or anomalous agent behaviour patterns
Detailed documentation for this investigation flow is available at:
Last updated