The AI governance and AI literacy glossary
500 terms, defined plainly. Every one is taught in more than one GAGE program, and most carry several definitions, because a word like deployer means something different to a lawyer, a data officer and an engineer. Each page shows all of them.
These are definitions, not lessons. Free to read, no account.
A
- A/B test
- A2A (Agent2Agent)
- Accountability Gap
- Accountability principle
- Action
- Action space
- Adequacy decision
- Adversarial manipulation
- Agent
- Agent (agentic AI)
- Agent (AI agent)
- Agentic AI3 programs
- Agentic AI value chain
- Agentic system
- Aggregation
- AI Agent
- AI Literacy (Article 3(56))
- AI Office5 programs
- AI Verify
- AI Verify Foundation
- AI-washing4 programs
- AIGP (Artificial Intelligence Governance Professional)
- Alarm fatigue4 programs
- Alert fatigue5 programs
- Algorithmic disgorgement3 programs
- Annex III
- Annex III (EU AI Act)3 programs
- Annex IV3 programs
- Annex VI (internal control)
- Annex XII
- Anonymization3 programs
- API (Application Programming Interface)3 programs
- Approval gate
- Article 4 (EU AI Act)
- Article 50 (transparency obligations)
- Article 6(3) filter
- Article 74 (market surveillance)
- Artifact chain
- As-of date3 programs
- ASEAN Digital Economy Framework Agreement (DEFA)
- ASEAN Secretariat
- Attribution3 programs
- Audit right3 programs
- Audit trail3 programs
- Automated Employment Decision Tool (AEDT)4 programs
- Automation Bias7 programs
- Automation complacency
- Autonomy4 programs
- Autonomy level
B
- Backdrivability
- Backlash
- Base Rate4 programs
- Baseline5 programs
- Benchmark
- Bias audit
- Binding constraint
- Biometric data
- Biometric identifier3 programs
- Blameless postmortem3 programs
- Blast Radius5 programs
- BLUF (Bottom Line Up Front)
- Board inspection
- Borrowed authority
- Briefcase backup (export)
- Build Log6 programs
- Business email compromise (BEC)
C
- C2PA (Coalition for Content Provenance and Authenticity)4 programs
- Cadence
- Calibrated trust3 programs
- Calibration5 programs
- Callback
- Caller ID spoofing
- Capability versus reliability
- Capstone3 programs
- Cascading failure3 programs
- CE marking
- Certification
- Chain of custody4 programs
- Change narrative
- Chargeback
- Chief AI Officer (CAIO)3 programs
- CI (continuous integration)
- Circuit breaker3 programs
- Claim
- Cognitive load
- Cold-read test
- Common-mode failure
- Compatibility test
- Compensating control
- Completeness3 programs
- Compliance
- Compliance calendar
- Compounding errors
- Computer-use agent3 programs
- Concession3 programs
- Confabulation
- Confidence Score
- Configuration3 programs
- Conformity assessment3 programs
- Conformity file
- Confused deputy
- Containment
- Contemporaneous record3 programs
- Content credentials
- Content provenance
- Context window5 programs
- Contributing factor
- Controller
- Conway's Law
- Corrective action3 programs
- Corrective action plan
- Counterfactual4 programs
- Counterfactual test
- Coupling
- Cross-border transfer mechanism
- Currency statement
- Cyberspace Administration of China (CAC)
D
- Dark data
- Dark pattern
- Data broker
- Data classification
- Data localization
- Data minimization4 programs
- Data poisoning6 programs
- Data Processing Agreement (DPA)4 programs
- Data Protection Impact Assessment (DPIA)6 programs
- Data provenance
- Data sovereignty
- Data subject
- Datasheet for a dataset
- Decision log
- Decision record4 programs
- Decision rights
- Deduplication
- Deepfake3 programs
- Default deny
- Defense in depth4 programs
- Deliberate practice3 programs
- Demo-to-deployment gap
- Deployer7 programs
- Deployer (Article 3(4))
- Deployer (EU AI Act)
- Deployment3 programs
- Design defect
- Desirable difficulty3 programs
- Difference-in-differences
- Differential Privacy4 programs
- Digital Omnibus
- Digital Omnibus on AI5 programs
- Digital Omnibus on AI (Regulation (EU) 2026/1744)3 programs
- Digital replica
- Digital twin3 programs
- Direct prompt injection4 programs
- Discovery (legal)
- Disparate impact4 programs
- Distribution shift5 programs
- Documentation debt
- Documentation gap
- Domain randomization
- DORA (Digital Operational Resilience Act)
- Dossier3 programs
- Downstream provider
- DPIA (Data Protection Impact Assessment)
- Drift6 programs
- Dunning-Kruger effect3 programs
- Duty cycle
- Duty of care4 programs
E
- EDPB (European Data Protection Board)
- Effect size
- Embedded AI3 programs
- Error budget3 programs
- Escalation ladder6 programs
- Escalation path7 programs
- Escalation Protocol3 programs
- Escalation trigger4 programs
- ESTABLISHED / EMERGING / SPECULATIVE4 programs
- EU AI Act5 programs
- EU AI Act (Regulation (EU) 2024/1689)4 programs
- EU declaration of conformity
- EU declaration of conformity (Article 47)
- European AI Office
- Evidence gap3 programs
- Evidence hierarchy
- Evidence index3 programs
- Evidence type
- Excessive agency3 programs
- Executive Order 141793 programs
- Expanded ASEAN Guide on AI Governance and Ethics, Generative AI
- Expected value
- Exposure3 programs
- Extraterritorial reach4 programs
F
- Fail-operational
- Failure Mode4 programs
- Failure Mode and Effects Analysis (FMEA)3 programs
- Fair Credit Reporting Act (FCRA)
- Fair use
- Feature flag3 programs
- Feature store
- Federal Trade Commission (FTC)
- Federated Learning3 programs
- Feedback loop4 programs
- Field checklist
- Finding3 programs
- Fine-Tuning5 programs
- Five Whys4 programs
- Force density
- Forgetting curve3 programs
- Foundation model3 programs
- Four-bar linkage
- Fundamental rights impact assessment (FRIA)
G
- Gap register4 programs
- GDPR (General Data Protection Regulation)4 programs
- GDPR Article 223 programs
- Gear ratio
- General-purpose AI (GPAI)3 programs
- General-Purpose AI (GPAI) Model
- General-purpose AI assistant
- General-purpose AI model (GPAI model)
- Generalization
- Generative AI3 programs
- Goodhart's Law3 programs
- Governance theater
- GPAI (general-purpose AI)4 programs
- GPAI Code of Practice
- Graceful degradation3 programs
- Gravity compensation
- Ground truth4 programs
- Grounding3 programs
H
- Hallucination4 programs
- Handoff
- Handover
- Happy path
- Hard case3 programs
- Hardware interlock
- Harmonised standard
- Harmonized standard3 programs
- High-risk AI system4 programs
- Hindsight bias
- Historical bias
- Holding torque
- Horizon scanning
- Human override rate
- Human-in-the-loop3 programs
- Human-in-the-Loop (HITL)3 programs
- Human-on-the-loop3 programs
- Human-on-the-Loop (HOTL)
- Hybrid architecture3 programs
- Hysteresis
I
- IMDA (Infocomm Media Development Authority)
- Impedance control
- Incident commander5 programs
- Incident log3 programs
- Incident response
- Indemnification3 programs
- Indemnity
- Independent verification4 programs
- Indirect prompt injection4 programs
- Inference4 programs
- Infocomm Media Development Authority (IMDA)3 programs
- Intended purpose
- Intended use
- Internal audit
- Internet Crime Complaint Center (IC3)
- ISO/IEC 420015 programs
- ISO/IEC 42001:20236 programs
- ISO/TS 15066:2016
J
- Jitter
- Judgment
- Judgment Layer3 programs
- Just culture
K
- Kill criteria4 programs
- Kill criterion3 programs
- Kill switch3 programs
L
- Lagging indicator4 programs
- Large Language Model (LLM)3 programs
- Latency budget
- Lawful basis
- Layering
- Leading indicator4 programs
- Least privilege6 programs
- Legal hold4 programs
- Legitimate interest
- Legitimate interest assessment (LIA)
- Liability cap
- Likelihood3 programs
- Living credential5 programs
- Living document
- Living Repository of AI Literacy Practices3 programs
- Load-bearing claim4 programs
- Loaded labor rate
- Lock-in
- Lost in the Middle
M
- Machine Unlearning3 programs
- Machinery Regulation (EU) 2023/1230
- Management review
- Mandatory subject of bargaining
- Market Surveillance Authority4 programs
- MAS AI Risk Management Guidelines (AIRG)3 programs
- Material change3 programs
- MCP (Model Context Protocol)3 programs
- Meaningful human oversight
- MECE (Mutually Exclusive, Collectively Exhaustive)
- Membership inference
- Membership Inference Attack3 programs
- Mitigation
- Model AI Governance Framework (MGF)3 programs
- Model AI Governance Framework for Agentic AI
- Model AI Governance Framework for Agentic AI (Agentic MGF)
- Model AI Governance Framework for Generative AI
- Model card6 programs
- Model Context Protocol (MCP)4 programs
- Model Drift4 programs
- Model provenance
- Model registry3 programs
- Model risk management (SR 11-7)
- Monetary Authority of Singapore (MAS)3 programs
- Moral crumple zone3 programs
- Morphological computation
- Multi-Agent System3 programs
N
- National market surveillance authority3 programs
- Near miss4 programs
- Net present value (NPV)3 programs
- NIS2 Directive
- NIST (National Institute of Standards and Technology)3 programs
- NIST AI Risk Management Framework (AI RMF)4 programs
- NIST AI RMF3 programs
- Node
- Normalization of deviance
- Notification clock
- Notified body5 programs
O
P
- Passkey
- Password manager
- Payback period4 programs
- PDPA (Personal Data Protection Act)
- Personal Data Protection Act (PDPA)
- Personal Data Protection Commission (PDPC)3 programs
- Phase gate
- Phishing
- Physics engine
- Pilot4 programs
- Pilot purgatory
- PIPA (Personal Information Protection Act)
- PIPC (Personal Information Protection Commission)
- PIPL (Personal Information Protection Law)
- Plan-Do-Check-Act (PDCA)
- Population Stability Index (PSI)
- Post-incident review
- Post-market monitoring
- Pre-mortem5 programs
- Pre-registration
- Predetermined Change Control Plan (PCCP)
- prEN 18286
- Preprint
- Presumption of conformity4 programs
- Primary source5 programs
- Profiling
- Profiling override
- Prompt injection7 programs
- Provenance4 programs
- Provider4 programs
- Provider (Article 3(3))
- Provider (EU AI Act)3 programs
- Proximate cause
- Pseudonymisation
- Psychological Safety3 programs
- Publicly Available Exception
- Pull request (PR)
- Purpose drift
- Purpose limitation4 programs
- Pyramid Principle3 programs
Q
R
- Race condition
- RAG (Retrieval-Augmented Generation)
- Rail
- Re-review trigger
- Reach3 programs
- Reasonably foreseeable misuse
- Recommendation
- Records of processing activities (ROPA)3 programs
- Recoverability
- Red team
- Red-teaming
- Redeployment3 programs
- Redundancy3 programs
- Regression
- Regulatory sandbox3 programs
- Reinforcement
- Reproducibility3 programs
- Residual Risk6 programs
- Reskilling
- Responsible disclosure
- Retention period4 programs
- Retrieval-Augmented Generation (RAG)6 programs
- Return on investment (ROI)4 programs
- Reverse image search
- Reversibility5 programs
- Review trigger3 programs
- Revocation3 programs
- Reward hacking
- Right of publicity4 programs
- Risk cost
- Risk register4 programs
- Risk tier
- Risk tiering4 programs
- Root cause5 programs
- ROS 2 (Robot Operating System 2)
- Rubber-Stamping3 programs
S
- Safe harbor
- Satisficing
- Scope boundary
- Scope creep4 programs
- Scoping
- Seam3 programs
- Self-attestation3 programs
- Self-disclosure
- Semantic search
- Sensitivity analysis3 programs
- Series elastic actuator
- Series elastic actuator (SEA)
- Service account
- Severity3 programs
- Shadow AI5 programs
- Shadow IT
- Silent failure4 programs
- Sim-to-real gap
- Single point of failure5 programs
- Social engineering
- Spacing effect3 programs
- Split-view poisoning
- Spoliation
- Staged rollout4 programs
- Standard Contractual Clauses (SCCs)3 programs
- Statistical significance
- Stop mechanism
- Stranger test3 programs
- Strict liability3 programs
- Structural control
- Structural fix
- Sub-processor4 programs
- Substantial modification
- Sunk cost3 programs
- Sunk cost fallacy
- Supervision tax
- Supervisory authority
- Surveillance audit
- Synthetic data5 programs
- System prompt
- Systemic risk
- Systemic-risk GPAI3 programs
T
- Tabletop exercise
- Tacit Knowledge3 programs
- TAKE IT DOWN Act
- Task Decomposition
- Technical debt4 programs
- Technical documentation
- Technical report
- Teleoperation
- Text and data mining (TDM) opt-out
- Threshold3 programs
- Timeliness3 programs
- Token3 programs
- Tokenization
- Tool
- Tool call
- Total cost of ownership (TCO)4 programs
- Trade secret3 programs
- TRAIGA (Texas Responsible Artificial Intelligence Governance Act)
- Training3 programs
- Training cutoff
- Training Data4 programs
- Training-serving skew
- Transfer
- Triangulation3 programs
- Trigger5 programs
- Trigger event
- Trust boundary3 programs