AI system

In short

An AI system is a machine-based system that works out from the input it receives how to produce outputs such as predictions, content, recommendations or decisions. It is the full working product, not just the model inside it.

Also called: Artificial intelligence system
Topic: The basics
Level: Beginner
First described: The OECD set out a definition in its 2019 Recommendation on AI, which NIST and others adapted⁴. OECD countries approved a revised version in 2023¹.

"AI system" is the term lawmakers and standards bodies use for AI as people actually meet it: a chatbot, a CV-screening tool, a fraud checker.

The most widely used definition comes from the OECD. Its 2023 update describes a machine-based system that, for explicit or implicit objectives, "infers, from the input it receives, how to generate outputs" such as predictions, content, recommendations or decisions, with varying levels of autonomy and adaptiveness¹. The EU AI Act's legal definition in Article 3(1) closely follows that wording². The US standards body NIST uses a similar definition, adapted from the OECD and ISO/IEC 22989⁴.

The key word is "infers". The EU AI Act says the definition should not cover software that simply runs rules written by people³. A spreadsheet formula is not an AI system; a tool that learned its own rules from data usually is.

A system is more than a model. The AI Act states that AI models "do not constitute AI systems on their own" and need extra parts, such as a user interface, to become one⁵.

The UK has no single legal definition. Its 2023 white paper instead defined AI by two traits, "adaptivity" and "autonomy", and left each regulator to apply them in its own sector⁶.

Example

ChatGPT, the app, is an AI system: it combines a model with an interface, safety filters and other software. The GPT model running inside it is one component, which the EU AI Act treats separately⁵.

Why it matters

Whether software counts as an "AI system" decides whether laws like the EU AI Act apply to it. The definition is the gateway to every other rule.

Questions people ask

What is the difference between an AI model and an AI system?

The model is the trained engine; the system is the full product built around it. The EU AI Act says models are essential components of AI systems but need further components, such as a user interface, to become systems⁵.

What counts as an AI system under the EU AI Act?

A machine-based system, operating with some autonomy, that infers from its input how to generate outputs such as predictions, content, recommendations or decisions². Software that only follows rules written by people is meant to fall outside³.

Does UK law define an AI system?

Not in one general law. The government's 2023 white paper defined AI by its "adaptivity" and "autonomy" and asked sector regulators to interpret this in their own areas⁶.

Related terms

Sources

Back to the glossary