AI Glossary (A–Z)

A standard that allows AI agents to communicate and coordinate with each other to complete tasks across systems without human intervention.

A system that uses generative AI to autonomously plan, decide, and take action towards a goal. 

The framework of policies, controls, and processes that guide how AI systems are designed, deployed, and managed. 

The end-to-end process of designing, deploying, operating, monitoring AI systems. 

The uncontrolled spread of AI tools and agents across an organization without clear visibility, governance, or ownership.

The maximum number of text a model can consider at once when generating a response. 

A system that enables users and AI applications to securely search across internal organizational data sources, such as documents, databases, and knowledge bases. 

Additional training of a pre-trained model using specialized data to improve performance for a specific task or domain. 

A large, general-purpose model trained on broad datasets that can be adapted to many use cases. 

AI systems that create new content such as text, images, code, audio, or video. 

Technical or policy-based constraints that limit AI behavior to reduce risk and misuse. 

When a generative AI system produces incorrect or fabricated information presented as fact. 

A governance approach where humans review, approve, or intervene in AI-generated outputs. 

The process where a trained AI model generates outputs in response to new inputs. 

A generative AI model trained on massive datasets to understand and generate Language. 

A method for defining what information, instructions, and constraints an AI model receives to ensure consistent and reliable outputs.

Performance degradation that occurs when a model’s outputs become less accurate over time. 

A centralized inventory of approved AI models, versions, and deployment metadata. 

A generative AI model capable of processing and generating multiple data types such as text, images, audio, or video. 

Human language, spoken or written, that AI systems can understand and generate to enable intuitive interaction.

The practice of tracking AI behavior, performance, and outputs over time. 

The input or instruction provided to a generative AI system that guides its output. 

The practice of designing prompts to improve accuracy, reliability, and safety of AI outputs. 

A security vulnerability where hidden or malicious instructions alter a model’s intended behavior. 

A structured testing approach where teams intentionally probe AI systems for weaknesses, failure modes, misuse scenarios, or security risks. 

An approach to building and using AI that emphasizes safety, transparency, accountability, and ethical use. 

A technique that allows generative AI systems to retrieve relevant external data before generating a response. 

A system that dynamically directs requests, tasks, or workloads to the appropriate AI model, agent, or resource based on rules, context, or policies. 

The continuous assessment and monitoring of an organization’s AI systems to identify security risks, misconfigurations, and policy violations. 

Unapproved or unmanaged use of AI tools or models within an organization. 

A lightweight language model optimized for efficiency and specific tasks, requiring fewer resources than large models. 

A high-level instruction that defines how a model or agent should behave across interactions. 

A basic unit of text (such as a word or word fragment) processed by a language model. 

A capability that enables AI models or agents to interact with external tools, APIs, or systems. 

A database optimized for storing and searching embeddings, often used in a retrieval-augmented generation system.