In short. A general-purpose AI model under the AI Act has broad task capabilities and can be integrated into varied systems, subject to the legal boundaries in Article 3.
Definition: Under Article 3(63) of the AI Act, a general-purpose AI model (GPAI model) displays significant generality, competently performs a wide range of distinct tasks and can be integrated into a variety of downstream systems or applications, regardless of how it is placed on the market.
Key facts
- The AI Act's recital 97 distinguishes models from AI systems: a model needs additional components, such as a user interface, to become a system.
- As of 17 October 2026, the Commission's scope guidelines provide a non-binding interpretation of the definition, including an indicative criterion based on training computation and output types.
- The Commission's implementation explanation dates model-provider obligations to 2 August 2025; providers of models placed on the market before that date have until 2 August 2027 to comply.
What separates a model from an answer engine?
The Act's Article 3(66) defines a general-purpose AI system as a system based on a general-purpose model that can serve a variety of purposes, directly or through integration into other systems. Applying this distinction to an answer engine requires examining the underlying model and the service that incorporates it separately.
The functional test appears in this extract from Article 3(63):
that displays significant generality and is capable of competently performing a wide range of distinct tasks regardless of the way the model is placed on the market
Recital 97 identifies application programming interfaces (APIs) and direct downloads as ways to place models on the market. The same recital explains that a provider integrating its own model into a marketed system remains subject to the model obligations.
How does the Commission assess general purpose?
The Commission's scope guidelines, paragraphs 15 to 20, use training compute, the computational resources used to train a model, measured in floating-point operations (FLOP). Their indicative criterion combines training compute greater than 10^23 FLOP with the ability to generate language as text or audio, images from text, or video from text.
The functional definition still governs exceptional cases. The guidelines explain that a model can meet the indicative criterion while lacking the required breadth of capabilities, or fall below it while possessing those capabilities. Their example of a model limited to transcribing speech illustrates the first situation. Paragraph 9 reserves authoritative interpretation of the Act to the Court of Justice of the European Union.
Why does the boundary matter for a website?
A provider is the person or organisation that develops a model, or has it developed, and places it on the market under its own name or trademark. The Act's Articles 2 and 3 cover providers placing models on the Union market, including providers established outside the Union. Applied to a website business, this calls for identifying its role in developing, marketing or integrating the model.
For publishers, Article 53 connects this classification to copyright and training transparency. It requires model providers to maintain a policy for compliance with Union copyright law, including identifying and complying with reservations of rights under Article 4(3) of Directive (EU) 2019/790. It also requires a publicly available, sufficiently detailed summary of training content using the AI Office template.
Related terms
FAQ
Does the definition cover a research prototype?
Article 3(63) excludes models used for research, development or prototyping before they are placed on the market. Recital 97 explains that subsequent market placement brings the relevant model obligations into play.
Is every general-purpose model a model with systemic risk?
Article 3(65) defines systemic risk as risk specific to high-impact capabilities, with significant Union-market impact, that can propagate at scale across the value chain; Article 3(64) defines those capabilities as matching or exceeding those recorded in the most advanced general-purpose models. Article 51 establishes an additional classification based on those capabilities or a Commission decision concerning equivalent capabilities or impact. Article 55 attaches additional obligations to that classification, including model evaluations and systemic-risk assessment and mitigation.
Does an open-source licence remove the provider obligations?
Article 53(2) exempts providers of qualifying free and open-source models from technical and downstream documentation duties. The licence must allow access, usage, modification and distribution, and the parameters, including weights, architecture information and usage information must be publicly available. The copyright-policy and training-summary duties remain, and models with systemic risk cannot use that exception.
Sources
- European Parliament and Council, "Regulation (EU) 2024/1689 (Artificial Intelligence Act)", 12 July 2024
- European Commission, "Commission Guidelines on the scope of the obligations for providers of general-purpose AI models established by Regulation (EU) 2024/1689 (AI Act)", 19 November 2025
- European Commission, "Learn more about the guidelines for providers of general-purpose AI models", 18 July 2025
Check how a page reads to answer engines. Open the GEO score analyzer