With the EU AI Act, the definition of a general-purpose AI system is provided within the framework:
‘general purpose AI system’ means an AI system that can be used in and adapted to a wide range of applications for which it was not intentionally and specifically designed; (Article 3)
This definition under the EU AI Act outlines that a general-purpose AI system is one that is capable of operating and adapting in various applications, even those beyond its initial design and intention. This means it’s not restricted to a specific, pre-defined role or function, but can be deployed in multiple scenarios and fulfill different tasks.
However, when considering the different aspects of the AI Act, such as Annex III and various articles (Article 5, Article 28a), the Act does not provide further details or specifications related to general-purpose AI systems. Annex III, which catalogues high-risk AI systems, does not explicitly reference this type of AI system. Additionally, Article 5 focuses on prohibited practices involving AI systems, and Article 28a on providers’ obligations, without particular references to general-purpose systems.
In order to ensure a level playing field and an effective protection of rights and freedoms of individuals across the Union and on international level, the rules established by this Regulation should apply to providers of AI systems in a non-discriminatory manner… (Recital 10)
The recitals, in general, provide an overarching view of the EU AI Act’s purpose and approach. Recital 6d describes the function and characteristics of AI systems but does not specifically reference a definition for general-purpose AI systems. Recital 10 emphasizes the application of the regulation to all AI system providers, illustrating the broad approach to regulation, which could be expected to encompass general-purpose AI systems as well.
In conclusion, general-purpose AI systems are defined directly in Article 3 of the EU AI Act. Given the broad applicability of the AI system under this definition, it can be inferred that the regulatory provisions and standards described throughout the Act would apply to these systems as part of the wider AI landscape, even if they are not always explicitly mentioned.
“‘general purpose AI system’ means an AI system that can be used in and adapted to a wide range of applications for which it was not intentionally and specifically designed” (Article 3)
The EU AI Act provides a definition of a general-purpose AI system in Article 3. According to the definition, a general-purpose AI system is an AI system that can be adapted and used in various applications, not bound by the specific intent or design it was originally created for. This suggests that the AI legislation acknowledges the versatility and adaptability of certain AI systems and deems it necessary to specifically define and regulate them.
The other articles (5 and 6) passed for analysis do not provide further insight or specify the context concerning a general-purpose AI system in the scope of this Act. Consequently, based on the analysis, the answer to the legal question “Does EU AI Act include a definition of a general-purpose AI system?” is affirmative. The definition is found in Article 3 of the Act.
‘general purpose AI system’ means an AI system that can be used in and adapted to a wide range of applications for which it was not intentionally and specifically designed; (Article 3)
This quote provides the definition of a general-purpose AI system under the AI Act. According to Article 3 of the Act, a general-purpose AI system refers to an AI system that is capable of being employed and modified for a broad spectrum of applications, even though it may not have been specifically and deliberately designed for these applications. This suggests that the Act recognizes the flexibility and versatility inherent in many AI systems, and seeks to ensure that such systems, irrespective of their originally intended functions, fall within the scope of the Act.
Therefore, the EU AI Act does include a definition of a general-purpose AI system as an AI system capable of being used in and adapted for a wide range of applications, even if it was not intentionally and specifically designed for those applications.
“‘general purpose AI system’ means an AI system that can be used in and adapted to a wide range of applications for which it was not intentionally and specifically designed;” (Article 3)
As defined in Article 3 of the EU AI Act, a ‘general purpose AI system’ is an AI system with the flexibility to be used and adapted for many different applications that go beyond its initially intended and specific design. This definition highlights the adaptability of these AI systems to handle a vast array of tasks, implying that they can function effectively in multiple domains.
”AI systems often have machine learning capacities that allow them to adapt and perform new tasks autonomously. Machine learning refers to the computational process of optimizing the parameters of a model from data, which is a mathematical construct generating an output based on input data. Machine learning approaches include, for instance, supervised, unsupervised and reinforcement learning, using a variety of methods including deep learning with neural networks. […] The function and outputs of many of these AI systems are based on abstract mathematical relationships that are difficult for humans to understand, monitor and trace back to specific inputs. These complex and opaque characteristics (black box element) impact accountability and explainability.” (Recital 6a)
In Recital 6a, it is suggested that AI systems can adapt over time and take on new tasks autonomously, a characteristic that might arguably link to the broad capabilities of general purpose AI systems. However, Recital 6a does not directly provide a definition of a ‘general-purpose AI system’.
”Foundation models are a recent development, in which AI models are developed from algorithms designed to optimize for generality and versatility of output. Those models are often trained on a broad range of data sources and large amounts of data to accomplish a wide range of downstream tasks, including some for which they were not specifically developed and trained. […] AI systems with specific intended purpose or general purpose AI systems can be an implementation of a foundation model, which means that each foundation model can be reused in countless downstream AI or general purpose AI systems. These models hold growing importance to many downstream applications and systems.” (Recital 60e)
Recital 60e provides some context to our understanding of ‘general purpose AI system’ by discussing ‘foundation models.’ These models are designed to optimize for generality and versatility, however, the recital does not provide a standalone definition for ‘general-purpose AI system’.
”Considering Annex III, it does not contain any definition or specific mention of a ‘general-purpose AI system’. Instead, it focuses on several types of high-risk AI systems subject to Article 6(2) of the Act. The AI systems described are categorized based on their application areas.” (Annex III)
A review of Annex III shows that it does not provide a definition for ‘general-purpose AI system’. The Annex specifically lists high-risk AI systems that are categorized based on their deployment in specific sectors.
In conclusion, while the EU AI Act doesn’t provide an exhaustive explanation of a general-purpose AI system beyond Article 3, it does offer a definition and indicates the broad and flexible character of such systems. Furthermore, context from several parts of the Act gives us an understanding of the characteristics that could possibly define a general-purpose AI system.