Glossary
Say AI assurance when you mean AI assurance.
Definitions this site uses so buyers, auditors, and assistants share the same language.
AI assurance
Independent, evidence-backed confidence that an AI system is the one you think it is, behaves within bounds, and can be shown to others.
AI governance
Accountability structures for AI — roles, policies, and decisions. Necessary, but not a substitute for testing the system.
Shadow AI
AI systems in use without a registered owner, intended use, or assurance record.
Agentic AI
Systems that plan, call tools, write memory, and loop toward a goal.
Model Context Protocol (MCP)
A protocol for exposing tools, resources, and prompts to models and agents.
AI bill of materials (AI BOM)
A point-in-time list of models, data, tools, and vendors that compose an AI system.
Explainable assurance score
A 0–100 score with named dimensions and listed deductions — not a mysterious trust badge.
EU AI Act
The European Union’s regulation on AI systems, with obligations that depend on use-case class.
ISO/IEC 42001
The AI management system standard.
NIST AI RMF
US National Institute of Standards and Technology AI Risk Management Framework: Govern, Map, Measure, Manage.
Prompt injection
An attack that causes a model or agent to follow hostile instructions inside content or tools.
Retrieval-augmented generation (RAG)
A pattern that retrieves documents or vectors before the model answers.
AI red teaming
Adversarial testing of models and agents to find harmful or insecure behavior.
Runtime enforcement
Blocking or allowing model/agent actions in the live request path.
AI assurance vs GRC
Assurance tests and evidences AI systems. GRC runs enterprise control programs.