AI Security Glossary

Search our knowledge base to break down complex AI terms and jargon into clear, practical definitions — many paired with short explainer videos for quick learning. Use it to understand key concepts, stay current with regulations, and navigate AI security with confidence.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Showing 00 of 00 terms

Term

Short Description

PointGuard

Video

Adversarial Machine Learning

A tactic where attackers craft inputs designed to fool AI models into making incorrect predictions. These subtle manipulations pose serious risks to model reliability, accuracy, and trust.

Protection

Agent Authentication

Agent authentication is the process of verifying the identity of an autonomous AI agent before it is allowed to act, call tools, or access data. It pairs with agent identity and agent authorization to form the identity triad for agentic systems.

Feature

Agent Authorization

Agent authorization is the process of deciding, at every action, whether an autonomous AI agent is permitted to invoke a given tool, access a given resource, or perform a given operation. It applies policy at the level of intent and tool call, not just the network or API surface.

Feature

Agent Covert Channel

An agent covert channel is an unapproved communication path that AI agents, or attackers using them, create through shared systems such as caches, repositories, or metadata, allowing data or instructions to move between sessions, agents, or accounts that should be isolated.

Protection

Agent Credential Theft

Agent credential theft is the stealing of tokens, API keys, and configuration files that AI agents use to access systems, often by infostealer malware, giving attackers the agent's access without needing to bypass multi-factor authentication.

Protection

Agent Discovery

Agent discovery is the practice of identifying every autonomous AI agent operating across an organization, including developer agents, customer-facing agents, internal automations, and third-party agents embedded in SaaS apps. It is the agentic counterpart to traditional asset discovery.

Feature

Agent Governance

Agent governance is the framework of policies, controls, and oversight that defines how autonomous AI agents are deployed, authorized, monitored, and retired. It extends traditional IT and AI governance to cover identity, authorization, behavior, and accountability for every agent in production.

Feature

Agent Hijacking

Agent hijacking is the adversarial takeover of an autonomous AI agent, in which an attacker gains control of the agent's reasoning, tools, or credentials. NIST treats agent hijacking as a distinct adversarial machine learning threat that covers a broader scope than goal-level redirection.

Protection

Agentic AI Security

Agentic AI Security is the discipline of protecting autonomous AI agents that perceive, plan, and act on enterprise systems. It combines identity, authorization, and runtime controls so agents can use tools, access data, and call APIs without exceeding their intended scope or trust boundary.

Feature

Agentic Application Security

The protection of autonomous AI systems—such as agents or AI-driven workflows—that independently perform tasks, access tools, or make decisions on behalf of users.

Feature

Agentic Browser

An agentic browser is a web browser with a built-in AI agent that can read pages, navigate, fill forms, and take actions for the user. Examples include Perplexity Comet and AI features in Chrome and Edge. Agentic browsers create new risks because the agent acts within the user's signed-in sessions.

Agentic Endpoint Security

Agentic endpoint security is the discovery, monitoring, and control of AI agents running on employee devices, including coding agents, agentic browsers, desktop assistants, and local MCP servers.

Feature

Agent Identity

Agent identity is the verifiable identifier and credential set assigned to an autonomous AI agent so its actions can be authenticated, authorized, and audited. Like user or service identities, agent identity binds an actor to permissions, sessions, and accountability across systems.

Feature

Agent Observability

Agent observability is the ability to trace and understand what AI agents are doing, including their prompts, reasoning steps, tool calls, data access, and outputs, across entire workflows. It is the foundation for detecting misuse, debugging failures, and proving compliance.

Feature

Agent Registry

An agent registry is an authoritative inventory of the AI agents operating in an organization, recording each agent's owner, purpose, identity, permissions, tools, and status. It is the system of record for agent governance.

Feature

Agent Sandbox

An agent sandbox is an isolated execution environment that confines what an autonomous AI agent can do, see, and reach. Sandboxing limits blast radius when prompt injection or misalignment causes the agent to attempt destructive or out-of-scope actions.

Feature

Agent Sprawl

Agent sprawl is the uncontrolled growth of AI agents across an organization's teams, platforms, and endpoints, creating overlapping, unowned, and poorly secured agents that security teams cannot see or manage.

Protection

Agent Swarm

An agent swarm is a large group of AI agents working in parallel and coordinating toward shared goals. Swarms can multiply productivity, but when they coordinate outside approved channels they can also amplify attacks, as in the 2026 OpenAI agent breach of Hugging Face.

Agent-to-Agent (A2A) Communication

Agent-to-agent communication describes the protocols and patterns by which autonomous AI agents exchange messages, tasks, and tool results with each other. As multi-agent architectures proliferate, A2A communication becomes both a productivity primitive and a fast-growing attack surface.

AI Agents / Agentic AI

Autonomous or semi-autonomous systems that use AI to perceive, reason, and act toward goals—often coordinating across tools, APIs, or environments to perform complex, multi-step tasks on behalf of users.

AI Anomaly Detection

The process of identifying unusual or unexpected behavior in AI systems, inputs, or outputs—often signaling potential security threats, performance issues, or compliance violations.

Feature

AI Application Cybersecurity (AIAC)

AI Application Cybersecurity (AIAC) protects AI-enabled applications from threats that target models, data pipelines, and integrations. It extends application security principles—like vulnerability management and runtime protection—to the unique risks of AI systems and LLM-powered apps.

Feature

AI Backdoor Attack

A backdoor attack on an AI model embeds a hidden trigger during training or fine-tuning so that the model behaves normally except when the trigger appears in input, at which point it produces attacker-chosen outputs. Backdoors are a foundational topic in adversarial ML research.

Protection

AI Bias

Systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group of users over others.

Protection

AI Bill of Materials (AI-BOM)

An AI Bill of Materials (AI-BOM) is a comprehensive, structured inventory documenting all components of an AI system—including datasets, models, software, hardware, dependencies, configurations, and processes—to ensure transparency, traceability, security, and regulatory compliance throughout the AI lifecycle.

Feature

AI Coding Agent Security

AI coding agent security is the practice of protecting software development environments from risks created by autonomous coding tools such as Claude Code, Codex, Cursor, and Copilot, which can read code, run commands, install packages, and access secrets.

Feature

AI Content Moderation

AI content moderation evaluates AI inputs and outputs for harmful, unsafe, or non-compliant content. It is a baseline trust-and-safety capability for chatbots, copilots, and agents, and it overlaps with broader AI guardrails and AI firewall categories.

AI Data Governance

Refers to the processes and policies that ensure data used in AI systems is high-quality, secure, and used ethically. It includes data lineage, access control, labeling practices, and compliance with regulations like GDPR, HIPAA, and industry-specific AI laws.

Feature

AI Data Loss Prevention (AI DLP)

AI Data Loss Prevention (AI DLP) is a category of runtime controls that detects and blocks sensitive data from leaving the enterprise through AI systems. It addresses both inbound (training and grounding) and outbound (prompts, responses, tool calls) flows that traditional DLP cannot inspect.

Feature

AI Detection & Response

AI Detection & Response secures AI in real time by continuously monitoring behavior, spotting anomalies or attacks, and automatically or manually responding—through actions like blocking, alerts, or adjustments—to maintain integrity, compliance, and safety in the face of evolving AI threats.

Feature

AI Discovery

AI Discovery is a structured process where artificial intelligence systems analyze data to identify valuable opportunities, patterns, or insights, enabling organizations to prioritize and implement AI-driven solutions aligned with business goals. It helps bridge the gap between AI potential and practical, strategic use cases.

Feature

AI Explainability

The degree to which humans can understand how an AI system makes decisions. Important for trust, transparency, debugging, and regulatory compliance—especially in sensitive or high-stakes applications.

AI Firewall

An AI firewall inspects traffic flowing into and out of AI systems, blocking malicious prompts, sensitive data leakage, and policy violations in real time. The term is used more broadly than LLM firewall and is increasingly applied to agentic and multimodal workloads.

Feature

AI Governance

A structured approach to overseeing the security, compliance, and responsible use of AI systems—ensuring that AI operations align with organizational policies, regulations, and ethical standards.

Feature

AI Guardrails

Controls and constraints built into AI systems to prevent harmful, unethical, or unintended outcomes. Guardrails enforce safe behavior during training, deployment, and real-time operation.

Feature

AI Hallucination

A phenomenon where AI systems, especially language models, generate confident but false, misleading, or fabricated information not grounded in reality or available data.

Protection

AI Impact Assessment

AI Impact Assessment is a critical process that evaluates the potential effects and implications of AI technologies on society, the environment, and the economy.

AI Incident Reporting

AI incident reporting is the practice, and increasingly the legal obligation, of notifying regulators, affected parties, or the public when an AI system causes or nearly causes serious harm, such as a security breach, safety failure, or loss of control.

AI Input Manipulation

A tactic where attackers or users intentionally craft inputs to alter or exploit an AI model’s behavior. Often used to bypass filters, trigger harmful outputs, or expose vulnerabilities.

Protection

AI Inventory

Comprehensive cataloging and management of all artificial intelligence systems, models, datasets, and related resources within an organization.

Feature

Video

AI Jailbreaking

A technique used to bypass restrictions or safety filters in AI systems—causing them to generate disallowed, dangerous, or otherwise restricted content.

Protection

AI Red Teaming

Security practice where experts simulate attacks on AI systems to identify vulnerabilities, assess their robustness, and improve their resilience against real-world threats.

Feature

Video

AI Resource Exhaustion Attacks

A denial-of-service tactic where attackers overload an AI system’s compute, memory, or token limits—causing slowdowns, failures, or forced shutdowns of models or APIs.

Protection

AI Runtime Protection

AI runtime protection, also called AI runtime security, is a security layer that monitors and defends AI systems while they operate, detecting threats, enforcing policies, and responding to misuse in real time.

Feature

AI Security Operations (AISecOps)

A security operations discipline focused on protecting AI systems from attacks, misuse, drift, and exposure—combining real-time monitoring, threat detection, and policy enforcement across the AI lifecycle.

Feature

AI Security Platform (AISP)

An AI security platform (AISP) is a single platform that protects an organization’s AI agents, models, applications, and data. It combines discovery, security testing, posture management, runtime guardrails, and governance, giving enterprises centralized visibility and control across the AI lifecycle.

Feature

AI Security Posture Management

Continuously monitoring, assessing, and improving the security measures and defenses of AI systems to ensure they are protected against threats and vulnerabilities.

Feature

Video

AI Security Severity Index (AISSI)

The AI Security Severity Index (AISSI) is a weighted scoring framework developed by PointGuard AI to track, categorize, and compare AI security incidents. It measures real-world severity, exploitability, and spread potential, providing a repeatable way to prioritize response and benchmark events.

Feature

AISI (AI Safety Institute)

AI Safety Institute (AISI) is the name used by several national bodies, including the UK and US AISIs, established to evaluate frontier model safety and inform government policy. AISI evaluations are increasingly referenced in CISO compliance discussions and procurement.

AI Supply Chain Attack

An AI supply chain attack targets the components that feed AI systems, including foundation models, training data, embeddings, plugins, MCP servers, and agent skills. Attackers compromise upstream artifacts so that downstream organizations unknowingly inherit the malicious behavior.

Protection

AI Supply Chain Security

AI supply chain security is the practice of protecting the models, datasets, libraries, MCP servers, and agent tools that AI systems depend on, so vulnerabilities and tampering are caught before they reach production.

Feature

Video

AI Threat Detection

AI threat detection identifies adversarial activity targeting AI systems, including prompt injection, model abuse, data exfiltration, and agent misuse. It pairs with AI Detection & Response (AIDR) as the runtime counterpart to traditional EDR and NDR in the AI stack.

Feature

AI Toxicity

The generation of harmful, offensive, or biased content by AI systems—often due to flawed training data, poor filtering, or misaligned model objectives.

Protection

AI Training Data

The datasets used to teach AI models how to recognize patterns, make decisions, or generate content. Training data quality and composition directly affect model performance, bias, and security.

AI TRiSM

A framework that combines Trust, Risk, and Security Management to govern AI systems—ensuring they operate responsibly, ethically, and within regulatory and organizational boundaries.

AIUC-1

AIUC-1 is a security, safety, and reliability standard and certification for AI agents. It defines controls that agent providers can be audited against, and its 2026 updates added requirements for MCP security and agent identity.

AI Usage Control (AIUC)

AI Usage Control (AIUC) governs how users, systems, and applications can access, interact with, and utilize AI models and data. It enforces dynamic policies to prevent misuse, data leakage, and unauthorized prompts—ensuring safe and compliant AI behavior across the enterprise.

Feature

AI Visibility

AI visibility is the broader CISO-facing capability of understanding how AI is being used, what risks it introduces, and where blind spots remain. It combines AI discovery, posture data, and behavioral telemetry into a single operational view.

Feature

AI Worm / LLM Worm

An AI worm (sometimes called an LLM worm) is a self-propagating malicious payload that spreads between AI agents or AI-augmented applications. Research demonstrations such as Morris II and Mini Shai-Hulud show how worms can ride on prompts, tool calls, and shared content rather than traditional executable code.

Protection

API Security

Protecting Application Programming Interfaces (APIs) from vulnerabilities, ensuring secure data exchange, and preventing unauthorized access or malicious activities.

Feature

Application Security Posture Management (ASPM)

Tools and processes for continuously monitoring, assessing, and improving the security posture of software applications throughout their development lifecycle, with a focus on identifying, assessing, and mitigating vulnerabilities and risks associated with applications to ensure they remain secure against potential cyber threats.

Feature

Video

Application Security Testing

Evaluating applications to identify and mitigate security vulnerabilities, ensuring that they are secure against threats and comply with security standards.

Artifact Repository

A storage location for software artifacts, including binaries and dependencies, used in the development and deployment process.

Artificial Intelligence (AI)

Technology that enables machines to perform tasks typically requiring human intelligence, such as reasoning, decision-making, and learning. Used in automation, analytics, and interaction across diverse business and consumer applications.

ASCII Smuggling

ASCII smuggling is a technique that hides text inside invisible Unicode characters, such as Unicode tag characters, so that AI models read hidden instructions or data that people cannot see. It is used for prompt injection and for covertly exfiltrating data.

Protection

Audit Trail

A chronological record of all activities and changes that have occurred in a system or process, ensuring transparency and accountability.

Automated Security Testing

The use of automated tools and processes to continuously test applications for security vulnerabilities throughout the development lifecycle.

Behavioral Trust Score (Agent)

A behavioral trust score is a dynamic, evidence-based rating of how closely an AI agent's observed behavior matches its declared intent and authorized policy. It augments static identity and authorization by adding a behavior-derived signal that can inform real-time authorization.

Feature

Blast Radius (AI Agent)

The blast radius of an AI agent is the maximum damage it could cause if compromised, manipulated, or malfunctioning, determined by the data, tools, credentials, and systems it can reach. Reducing blast radius is a primary goal of agent security design.

Ready to get started?

Our expert team can assess your needs, show you a live demo, and recommend a solution that will save you time and money.