Home / Glossary
AI security glossary: the terms security teams need to know
Plain-English definitions for the terms that came with employee AI: AI Usage Control (AIUC), shadow AI, MCP servers, AI agents and AI security platforms. Each page says what the term means, why it matters on the employee device and what a security team can do about it.
The category
AI Usage Control
What is AI Usage Control (AIUC)?
Control over which AI employees use and what each one may do, enforced on the device.
AI security platform
What is an AI security platform?
A platform that finds the AI an enterprise runs, stops the high-risk ones and keeps the evidence.
Enterprise AI security
What is enterprise AI security?
Protecting a company from the AI it uses, on the network, in the cloud and on devices.
Generative AI security
What is generative AI security?
How a company protects its data and systems when employees use generative AI.
Agentic AI security
What is agentic AI security?
Security for AI that plans and acts on its own, focused on reach and actions.
AI runtime security
What is AI runtime security?
Controlling AI while it runs, one action at a time.
Endpoint AI security
What is endpoint AI security?
Finding and controlling the AI that runs on employee devices.
Allow, ask or deny
Allow, ask or deny: how policy for AI agents works
The three verdicts a rule can give each action an AI agent tries.
Platform
What is an AI security platform at the endpoint?
A platform that works on the employee device to find AI, stop the risky ones and keep evidence.
Category
What is an Enterprise AI Security Platform?
The category AIBound sits in: AI security at scale, with central rules and evidence.
Endpoint AI
What is AI on the endpoint?
AI that runs on the employee device, where gateways and cloud tools cannot see it.
Deployment
What is agentless AI security?
AI security that connects through the tools you already run, not a new agent for every system.
Real time
What is real-time AI security?
Finding and stopping risky AI as it runs, not after a report or an audit.
Risk
What is high-risk AI?
AI that can expose data or act unchecked, and how that differs from the EU AI Act category.
Shadow AI and the workplace
Shadow AI
What is shadow AI?
AI your employees use that security never approved, from chatbots to coding agents.
AI gateway
What is an AI gateway?
A proxy between an app and an AI model that routes, logs and checks the traffic.
AI acceptable use policy
AI acceptable use policy: what it is and what to include
The written rules for AI at work. It tells employees, it does not show what runs.
AI inventory
What is an AI inventory?
The list of the AI in use. A list alone stops nothing, so action follows.
AI risk grading
What is AI risk grading?
A grade for each AI, so the high-risk ones are stopped first.
AI discovery
What is AI discovery?
Finding the AI your employees run, so the high-risk ones can be stopped.
Local AI
What are local AI models and local AI agents?
Models and agents that run on a device and may never touch a gateway.
AI extensions
What are AI extensions and plugins?
Add-ons that put AI inside browsers and code editors, installed in a click.
AI at work
What is AI in the workplace?
The AI tools employees use for their jobs, approved by the company or not.
Shadow agents
What are shadow agents?
AI agents your employees run that security never approved.
AI BOM
What is an AI BOM (AI bill of materials)?
The formal list of the models, data and software inside an AI system.
AI sprawl
What is AI sprawl?
AI tools and agents multiplying across a company faster than anyone tracks them.
AI agents
AI agent security
What is AI agent security?
How to control what AI agents can reach and do on employee devices.
AI identity security
What is AI identity security?
Which identity each AI agent runs as, and what that identity can reach.
AI Kill Switch
What is an AI kill switch?
One action that stops a rogue AI agent on every device at once.
AI guardrails
What are AI guardrails?
Rules that decide what an AI agent may do, checked on every action.
Personal AI agents
What is personal AI agent security?
Agents employees set up for themselves, what they reach and how to control them.
Coding agents
What is coding agent security?
Agents that write and run code on developer devices, and how to control them.
AI agent observability
What is AI agent observability?
Seeing what AI agents do, step by step. Control is a separate job.
AI agent governance
What is AI agent governance?
The rules, owners and evidence for the AI agents a company runs.
Least privilege for agents
Least privilege for AI agents
Give each AI agent only the access its task needs, and nothing more.
AI agents
What is an AI agent?
Software that uses an AI model to plan steps and act on its own.
Agentic AI
What is agentic AI?
AI that plans and acts toward a goal with little human direction.
AI agents
What are AI coding agents?
AI tools that write, run and change code on their own, and what they can reach.
Human in the loop
What is human in the loop?
A person approves an AI action before it happens.
Agent swarm
What is an agent swarm?
Many AI agents working together on one goal, often started by one person.
Agentic browser
What is an agentic browser?
A browser with an AI agent that clicks, types and acts for the user.
Computer use agent
What is a computer use agent?
An AI agent that works a computer through the screen, mouse and keyboard.
MCP and skills
MCP security
What is MCP security?
How to find and control the MCP servers that connect AI agents to your systems.
MCP server
What is an MCP server?
The small program that lets an AI agent reach tools, files and company systems.
Agent skills
What are agent skills?
Instruction files an AI agent loads to do a task, and why each one needs a check.
MCP gateway
What is an MCP gateway?
A central checkpoint for MCP traffic, and the servers on the device it cannot see.
OWASP MCP Top 10
OWASP MCP Top 10 and the Agentic Top 10
Two OWASP lists that rank the main risks of MCP servers and AI agents.
Shadow MCP server
What is a shadow MCP server?
An MCP server an employee added that security never reviewed.
MCP gateway vs endpoint
MCP gateway vs endpoint MCP control
A gateway controls routed MCP servers. Endpoint control finds the ones on devices.
Protocols
What is MCP (Model Context Protocol)?
The open standard that lets AI agents connect to tools and data through servers.
A2A protocol
What is the A2A (Agent2Agent) protocol?
The open standard that lets AI agents talk to each other and share tasks.
MCP authentication
What is MCP authentication?
How an MCP client proves who it is to an MCP server.
Governance and risk
AI governance
AI Usage Control (AIUC) vs AI governance
Governance writes the rules. AI Usage Control (AIUC) enforces them where the AI runs.
AI governance
What is AI governance?
The rules, roles and oversight that decide how a company may use AI.
AI governance framework
AI governance framework: what it is and what goes in one
The parts of a framework for governing AI, with three examples to build on.
ISO/IEC 42001
What is ISO/IEC 42001?
The international standard for running an AI management system.
NIST AI RMF
What is the NIST AI RMF?
A voluntary US framework for managing AI risk, built on four functions.
EU AI Act
EU AI Act compliance checklist: seven steps, and the dates that matter
Seven steps for the AI your employees use, with the dates as they stand today.
AI TPRM
What is AI TPRM (third-party risk management)?
How to assess AI vendors and the AI built into the products you already buy.
AI risk management
What is AI risk management?
How to find the AI you run, judge each tool and act on the risky ones.
OWASP LLM Top 10
OWASP Top 10 for LLM applications
The ten security risks OWASP ranks highest for apps built on LLMs.
AI risk assessment
What is an AI risk assessment?
The review of what an AI system could harm and what to do about it.
AI audit
What is an AI audit?
An independent check of how a company uses and controls AI.
AI TRiSM
What is AI TRiSM?
Gartner's term for AI trust, risk and security management.
Threats and attacks
Prompt injection
What is prompt injection?
Crafted text that makes an AI model follow an attacker's instructions.
AI data leakage
What is AI data leakage?
Sensitive company data leaving through AI tools, prompts and agents.
AI red teaming
What is AI red teaming?
Attacking an AI system on purpose to find its weak points before someone else does.
Indirect prompt injection
What is indirect prompt injection?
Hidden instructions in a page, file or email that an AI agent reads and obeys.
AI supply chain security
What is AI supply chain security?
Checking the models, skills, MCP servers and extensions that come from outside.
Excessive agency
What is excessive agency?
The OWASP risk of an AI agent that can do more than its task needs.
Tool poisoning
What is tool poisoning?
Malicious instructions hidden in the description of a tool an AI agent trusts.
Rogue AI
What is rogue AI?
An AI system or agent acting outside what its owner intended.
AI jailbreak
What is an AI jailbreak?
A prompt that gets an AI model to ignore the safety rules it was built with.
Other security tools
AI firewall
What is an AI firewall?
A filter on prompts and responses in the path between an app and a model.
AI-SPM
What is AI-SPM (AI security posture management)?
Posture management for the AI in cloud and SaaS accounts.
LLM security
What is LLM security?
Protecting large language models and the apps built on them from attack.
AI DLP
What is AI DLP?
Data loss prevention for prompts, AI tools and AI agents.
AIUC vs AI-SPM
AI Usage Control (AIUC) vs AI-SPM
AI-SPM checks the AI you build. AI Usage Control (AIUC) stops risky AI on devices.
AIUC vs DLP
AI Usage Control (AIUC) vs DLP
DLP watches data leaving. AI Usage Control (AIUC) controls which AI runs.
AIUC vs CASB
AI Usage Control (AIUC) vs CASB
A CASB reads traffic to cloud apps. AI Usage Control (AIUC) reads the device.
AIUC vs AI gateway
AI Usage Control (AIUC) vs an AI gateway
A gateway reads the traffic sent through it. Local AI may never pass through.
EDR vs endpoint AI
EDR vs endpoint AI security
EDR reads processes for malware. Endpoint AI security reads the AI itself.
EDR
What is EDR (endpoint detection and response)?
Software that watches each device for attacks and helps the security team respond.
DLP
What is DLP (data loss prevention)?
Tools and rules that find sensitive data and stop it leaving the company.
CASB
What is a CASB (cloud access security broker)?
A checkpoint between employees and cloud apps that shows usage and applies rules.
AI sandbox
What is an AI sandbox?
An isolated place to run an AI agent so it cannot reach real systems.
FAQ
Frequently asked questions
What is in the AI security glossary?
The AI security glossary explains the terms security teams meet as employees adopt AI: AI Usage Control (AIUC), shadow AI, MCP security, AI agent security and AI security platforms. Each entry is a short, plain-English page that defines the term, shows where it appears on the employee device and links to the matching AIBound page.
What is AI Usage Control (AIUC) in one sentence?
AI Usage Control (AIUC) is control over which AI your employees use and what each one may do, enforced on the endpoint where the AI runs. It covers agents, MCP servers, skills, models, apps and prompts, and it is the job AIBound does as an Enterprise AI Security Platform.
Which term should a new security team learn first?
Start with shadow AI, because you cannot control AI you cannot see. Then read what an AI agent can reach and what MCP security covers, because agents and MCP servers are where the reach and the risk sit. AI Usage Control (AIUC) ties the three together into one control on the device.
How do these terms connect to AIBound?
Each term maps to something AIBound does. AISeeker finds shadow AI, agents and MCP servers on employee devices. AgentTracer, SkillGuard and AgentHarness stop the high-risk ones in real time. AIAssure keeps the evidence for your GRC program. Every glossary page links to the product or solution page that matches its term.