General-purpose AI model

In short

A general-purpose AI model is the EU AI Act's legal term for an AI model that can do a wide range of different tasks and be built into many other products, such as the large models behind ChatGPT, Claude and Gemini.

Also called: GPAI model, GPAI, General purpose AI model, GPAI model with systemic risk
Topic: Law, policy and the UK
Level: Some background helps
First described: Defined in Article 3(63) of the EU AI Act, Regulation (EU) 2024/1689, published in July 2024¹.

The term comes from the EU AI Act. It defines a general-purpose AI model as one that shows significant generality, can competently perform a wide range of distinct tasks, and can be integrated into a variety of downstream systems or applications. Such models are often trained on large amounts of data using self-supervision at scale. Models used only for research, development or prototyping before release are excluded¹.

The model is not the same as the app. The Act separately defines a general-purpose AI system as an AI system based on such a model that can serve many purposes¹. Companies that have signed the EU's code of practice for these models include OpenAI, Anthropic, Google, Microsoft and Mistral AI⁶.

Every provider must keep technical documentation, give information to companies building on the model, have a policy to comply with EU copyright law, including opt-outs from text and data mining, and publish a sufficiently detailed summary of the content used for training³. Models released under a free and open-source licence with public weights are spared the two documentation duties, unless they carry systemic risk³.

A stricter tier covers models "with systemic risk". A model is presumed to qualify when the compute used to train it exceeds 10^25 floating point operations, and the European Commission can also designate models². Their providers must also evaluate the model, including adversarial testing, and assess and mitigate systemic risks at EU level⁴.

These rules have applied since 2 August 2025⁵. The Commission can fine providers up to 3% of worldwide annual turnover or EUR 15 million, whichever is higher⁷. The 2026 Digital Omnibus delayed the high-risk rules but left the model rules on their original schedule⁸.

Example

A start-up in Leeds that builds a customer-service chatbot on top of a model from a large lab is relying on a general-purpose AI model. The lab must give the start-up enough documentation to understand the model's capabilities and limits³.

Why it matters

It is the category that brings the biggest AI labs under EU law directly, including duties on training data summaries and copyright. If you use a mainstream chatbot, the model under it is probably covered.

Questions people ask

Is ChatGPT a general-purpose AI model?

The Act does not list products by name. But the models behind chatbots like ChatGPT are built to handle a wide range of tasks and are integrated into many apps, which is the Act's definition¹. OpenAI has signed the EU's code of practice for providers of these models⁶.

What is a general-purpose AI model with systemic risk?

It is the most capable tier. A model is presumed to have high-impact capabilities, and so systemic risk, when more than 10^25 floating point operations were used to train it. The European Commission can also designate a model after assessing it². These providers face extra duties such as adversarial testing⁴.

Do open-source models have to follow the rules?

Partly. Models released under a free and open-source licence with public weights do not have to produce the technical documentation for regulators and downstream developers. They still need a copyright policy and a training-content summary, and the exemption does not apply to models with systemic risk³.

Is the code of practice compulsory?

No. The European Commission calls it a voluntary tool that helps providers show they meet the Act's obligations. Signing it is meant to reduce paperwork and give more legal certainty⁶.

Related terms

Tools that use it

Sources

Back to the glossary