What is GPAI and what does it mean for fashion?
General-purpose AI: under the EU AI Act, models that can perform a wide range of tasks and be built into many systems.
In short
GPAI stands for general-purpose AI. Under the EU AI Act, it refers to models that can perform a wide range of tasks and be built into many different systems, such as large language models. Fashion companies using these models through an API are usually deployers with lighter duties than the model providers.
How does it work in practice?
A general-purpose model is not built for one task. The same model can write product descriptions, summarise retailer feedback, translate line sheets or power a B2B assistant. The EU AI Act places specific obligations on the providers of such models, including technical documentation, information for downstream developers and transparency about training content. Models considered to pose systemic risk face additional requirements.
Fashion companies typically meet GPAI in three ways:
- Using a general-purpose model through an API inside their own tools.
- Using software from vendors that embed such models.
- Fine-tuning a model on their own data for a specific task.
Why does it matter?
Understanding the role a company plays determines its obligations. A brand that simply uses a general-purpose model through an API is usually a deployer rather than a provider, with lighter duties. However, fine-tuning or substantially modifying a model, or placing it on the market under its own name, can change that role and bring provider obligations.
How does AI use it?
GPAI models are the engines behind most generative AI applications in fashion, from content creation to conversational commerce. They are often called foundation models, because other systems are built on top of them.
Common pitfalls
Companies sometimes assume that GPAI rules concern only large technology providers and ignore their own use. Others fine-tune models without checking how this affects their regulatory role. Contracts with vendors should clarify responsibilities, and internal guidelines should define which models can be used with which data. Documenting each use, and whether any customisation happened, keeps the company ready for questions from partners or regulators.
Frequently asked questions
Are large language models considered GPAI?
Large language models that can handle many different tasks are typical examples of general-purpose AI models. Their providers carry the main GPAI obligations under the EU AI Act.
Does fine-tuning a model make a company a GPAI provider?
It can, depending on how substantial the modification is and how the result is used or distributed. Companies planning significant fine-tuning should check their role with legal advisers.