Glossary

AI governance, defined plainly

Short, practical definitions of the terms that come up when you discover AI use, secure agents, and prove what happened.

A

Agent identityAgent identity is a distinct identity assigned to an AI agent so its access, credentials, tools, and permitted actions can be governed.
Agentic AIAgentic AI describes AI systems that can plan steps or use tools to pursue a task with some degree of autonomy.
AI agentAn AI agent is software that uses a model to carry out a task and may call tools or connected applications within its configured permissions.
AI audit trailAn AI audit trail is a reviewable record of AI-related decisions, approvals, denials, costs, and configuration changes.
AI bill of materialsAn AI bill of materials is a structured record of AI-related components or applications that an organization can review and share.
AI control planeAn AI control plane coordinates identity, policy, access, and evidence for AI systems and agents across supported environments.
AI cost governanceAI cost governance combines budgets, cost visibility, and usage decisions to help organizations manage spending on AI models and agents.
AI data loss preventionAI data loss prevention applies checks to AI traffic to identify or limit the exposure of sensitive information in supported paths.
AI gatewayAn AI gateway mediates traffic between applications or agents and AI services so configured controls can be applied along the request path.
AI governance platformAn AI governance platform helps an organization discover AI use, apply controls to supported AI activity, and retain reviewable evidence.
AI guardrailsAI guardrails are policies and runtime checks that guide model requests, responses, or agent actions within defined boundaries.
AI inventoryAn AI inventory is a record of detected AI applications and related details that help an organization review and classify AI use.
AI risk classificationAI risk classification is a repeatable way to categorize AI systems or uses and guide review according to organizational criteria.
AI runtime securityAI runtime security applies controls while AI agents or model requests execute, including identity, policy, network, credential, and action boundaries.

B

Bring your own keyBring your own key is a model access arrangement where an organization provides its own provider credential for use through a supported gateway.

C

Credential brokeringCredential brokering is a pattern where a gateway keeps a secret and provides an agent with scoped access instead of exposing the raw credential.

D

Default-deny egressDefault-deny egress is a network policy that blocks outbound connections unless an explicit rule allows a destination.

E

EU AI ActThe EU AI Act is a European Union regulation that sets obligations for providers and deployers of AI systems, with requirements phasing in over time.

H

Human in the loopHuman in the loop is an approach that gives a person a review or approval role in an AI workflow or action.

I

ISO/IEC 42001ISO/IEC 42001 is an AI management system standard that organizations can use to establish and improve AI governance processes.

M

N

NIST AI RMFThe NIST AI Risk Management Framework is a resource for organizing AI risk work into Govern, Map, Measure, and Manage functions.

P

S

Shadow AIShadow AI is AI use that an organization has not approved, cataloged, or fully reviewed through its governance process.

Put a halo around the AI your business runs.

See discovery, runtime policy, approvals, and evidence working together on your own use cases.

Book a walkthrough Start with Halo