Skip to main content

Glossary — AI Product

This glossary provides definitions of key terms used in the AI Product Specification (AIPS).
It focuses on AI-specific vocabulary and complements the site-wide glossary.

AI Asset​

A building block in the AI lifecycle — such as datasets, pretrained weights, embeddings, prompts, feature stores, or rulesets.
AI Assets are valuable, but not packaged as products unless they implement all true AI Product characteristics.

AI Product​

A packaged, consumable unit of AI capability that inherits the characteristics of a Product (as per BPS) and extends them with AI-specific attributes such as capability type, fairness, risk classification, and explainability.
See Definition.

Capability Type​

The declared category of capability provided by an AI Product.
Examples: Model, Agent, Multi-Agent System, Hybrid, AGI, ASI.
See Capability Types.

Productness​

The condition of satisfying all required characteristics of a true AI Product.
If the characteristics are missing, the entity is only an AI Asset.

Principle: Productness is earned, not assumed.

Prohibited Uses​

Explicitly declared forbidden applications of an AI Product (e.g., biometric surveillance, misinformation generation, autonomous weapons).
Ensures consumers do not misuse AI Products outside of governance boundaries.

Risk Classification​

The risk level of an AI Product, based on governance frameworks (e.g., EU AI Act categories: minimal, limited, high, unacceptable).
Provides consumers with awareness of ethical, legal, and societal implications.

Trustworthy Discovery Mechanism​

A catalog, registry, or marketplace that enables consumers to discover AI Products with confidence in the accuracy, governance, and reliability of the information.
See Glossary — Reference.

Drift​

The degradation of an AI Product’s performance over time due to changes in data, environment, or usage context.
Requires continuous monitoring and possible retraining.

Explainability​

The ability of an AI Product to provide transparent reasoning signals about its outputs.
Examples: feature importance, attention maps, natural language rationales, or system cards.

Autonomy​

The degree of independent action an AI Product (e.g., an agent) can take.
Levels may range from advisory to fully autonomous.
Declaring autonomy level is required for governance and consumer trust.

Summary​

  • AI Assets = building blocks.
  • AI Products = governed, consumable units of capability.
  • Productness defines the line between the two.
  • Capability Types, Risk, Prohibited Uses, Explainability, and Drift are unique attributes of AI Products.
  • Discoverability requires trustworthy discovery mechanisms.