01
Adoption
Definition
How often and how successfully users start using a product or feature.
Back to Agentic AI Glossary
Agentic AI Glossary
AI Safety, Ethics & Risk
AI Safety, Ethics & Risk terms and explanations from the Agentic AI Glossary.
64 terms in this chapter
02
Adoption Metrics
Definition
Measurements showing usage, activation, retention, acceptance, and engagement with an AI system.
03
AI Assistant
Definition
An AI experience that helps users complete tasks through conversation or guided actions.
04
AI Copilot
Definition
An AI assistant that works alongside a user instead of fully replacing the user.
05
AI Product
Definition
A product that uses AI to create value for users or automate part of a workflow.
06
Build vs. Buy
Definition
The strategic choice between developing an AI capability internally or purchasing a platform or vendor solution.
07
Business Value
Definition
The measurable benefit a product creates, such as revenue, savings, speed, or quality.
08
Change Management
Definition
Preparing people, processes, communication, and governance so teams adopt AI safely.
09
Citation
Definition
A reference to the source used to support an AI answer.
10
Confidence Score
Definition
A number or label that estimates how reliable an answer, prediction, or decision is.
11
Confirmation Step
Definition
A user approval step before an AI system performs an important or risky action.
12
Enablement
Definition
Training and supporting users, operators, developers, and stakeholders to work effectively with AI systems.
13
Feedback Button
Definition
A UI control that lets users rate, correct, or comment on an AI response.
14
High-Risk Action
Definition
An action that can cause meaningful harm, loss, compliance issues, or security exposure.
15
Human Approval
Definition
A design pattern where a person approves an AI action before it is completed.
16
Low-Risk Action
Definition
An action with limited impact that can usually be automated with fewer approvals.
17
Maturity Model
Definition
A staged framework for assessing current AI capability and planning improvement.
18
Onboarding
Definition
The process of helping users understand and start using an AI product effectively.
19
Personalization
Definition
Adapting an AI experience based on user context, history, role, preferences, or account data.
20
Risky Action
Definition
An AI action that should be checked because it may affect money, data, safety, or access.
21
ROI
Definition
Return on investment: value gained relative to cost.
22
Source Attribution
Definition
Showing where information came from so users can verify the answer.
23
Time to Value
Definition
The time required before users or the organization experience measurable benefit.
24
Transparency
Definition
Making AI behavior, limitations, sources, and actions understandable to users.
25
Trust
Definition
User confidence that the AI system is useful, safe, honest, and controllable.
26
Undo Action
Definition
A way for users to reverse or recover from an AI action.
27
User Control
Definition
Giving users clear choices, approvals, limits, and ways to correct the AI system.
28
User Goal
Definition
The outcome the user wants to achieve.
29
User Intent
Definition
The meaning or purpose behind a user request.
30
User Journey
Definition
The path a user follows from need to successful outcome.
31
User Preference
Definition
A user setting or habit that helps personalize the AI experience.
32
ABAC
Definition
Attribute-based access control, where permissions depend on attributes such as user role, resource type, location, or risk.
33
Abuse Testing
Definition
Abuse Testing validates abuse behavior through examples, scenarios, checks, or automated test runs. It helps catch issues before production.
34
Access Control
Definition
Permission management that determines who or what can access data, tools, actions, and administrative functions.
35
AI Red Teaming
Definition
AI Red Teaming describes a safety or security concern around ai red teaming. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
36
AI Safety
Definition
AI Safety describes a safety or security concern around ai safety. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
37
AI Security
Definition
AI Security describes a safety or security concern around ai security. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
38
Compliance
Definition
Conformance with legal, regulatory, contractual, and organizational requirements.
39
Credential Leakage
Definition
Credential Leakage describes a safety or security concern around credential leakage. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
40
Data Exfiltration
Definition
Data Exfiltration describes a safety or security concern around data exfiltration. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
41
Data Privacy
Definition
Protection of personal, sensitive, or confidential data throughout AI processing and storage.
42
Excessive Agency
Definition
Excessive Agency describes a safety or security concern around excessive agency. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
43
Explainability
Definition
The ability to describe why an agent produced an output or chose an action in terms understandable to humans.
44
Governance
Definition
Policies, ownership, processes, and controls that make AI systems accountable and responsible.
45
Indirect Prompt Injection
Definition
Indirect Prompt Injection describes a safety or security concern around indirect prompt injection. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
46
Insecure Output Handling
Definition
Insecure Output Handling describes a safety or security concern around insecure output handling. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
47
Jailbreak
Definition
Jailbreak describes a safety or security concern around jailbreak. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
48
Least Privilege
Definition
Least Privilege describes a safety or security concern around least privilege. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
49
LLM Security
Definition
LLM Security describes a safety or security concern around llm security. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
50
Model Denial of Service
Definition
Model Denial of Service describes a safety or security concern around model denial of service. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
51
Policy Violation
Definition
Policy Violation describes a safety or security concern around policy violation. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
52
Privilege Escalation
Definition
Privilege Escalation describes a safety or security concern around privilege escalation. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
53
RAG Poisoning
Definition
RAG Poisoning describes a safety or security concern around rag poisoning. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
54
RBAC
Definition
Role-based access control, where permissions are assigned to roles and users inherit access from their role.
55
Red Teaming
Definition
Red Teaming describes a safety or security concern around red teaming. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
56
Sandboxing
Definition
Sandboxing describes a safety or security concern around sandboxing. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
57
Secrets Leakage
Definition
Secrets Leakage describes a safety or security concern around secrets leakage. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
58
Secure Logging
Definition
Secure Logging describes a safety or security concern around secure logging. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
59
Security
Definition
Controls that protect AI systems, data, models, tools, and users from misuse or unauthorized access.
60
Supply Chain Vulnerability
Definition
Supply Chain Vulnerability describes a safety or security concern around supply chain vulnerability. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
61
Threat Modeling
Definition
Threat Modeling describes a safety or security concern around threat modeling. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
62
Tool Misuse
Definition
Tool Misuse describes a safety or security concern around tool misuse. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
63
Training Data Poisoning
Definition
Training Data Poisoning describes a safety or security concern around training data poisoning. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
64
Unauthorized Tool Access
Definition
Unauthorized Tool Access describes a safety or security concern around unauthorized tool access. Teams use it to reduce misuse, data exposure, unauthorized access, or harmful AI behavior.
Explore more chapters or test your knowledge with quizzes.