General-Purpose AI
AI capable of performing a wide range of tasks across domains.
How the laws define this
Regulatory frameworks align on the core idea that general-purpose AI operates across multiple domains and tasks rather than being narrowly specialized, possessing broad adaptability that triggers distinct risk management, governance, or regulatory scrutiny. The primary point of divergence lies in technical scope and terminology. Strategia Italiana per l'Intelligenza Artificiale 2024-2026 explicitly equates general-purpose AI with foundation models, defining them as large-scale models trained on broad datasets for downstream adaptation. In contrast, Draft Law B154/2024 — Proposal regarding Artificial Intelligence (Proiect de lege privind Inteligența Artificială, B154/2024), the National Artificial Intelligence Policy Framework (Draft) — Department of Communications and Digital Technologies, and the Strategy for the Development of Artificial Intelligence Technologies until 2030 characterize general-purpose AI more broadly by its multi-domain capabilities, systemic scale, or flexible reasoning and learning abilities, without strictly defining it as a dataset-trained foundation model.
Synthesised from the 4 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first. Showing 3 of the 4 instruments that define this term; all are listed under Definitions below.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| Uzbekistan | Strategy for the Development of Artificial Intelligence Technologies until 2030 | In Force | Emphasizes advanced systems with flexible reasoning and learning abilities, contrasting them explicitly with narrow AI in the context of long-term planning and infrastructure. |
| Romania | Draft Law regarding Artificial Intelligence, B154/2024 | Under Review | Highlights potential for adaptation across diverse applications, linking this capability to specific regulatory considerations unlike the others. |
| South Africa | National Artificial Intelligence Policy Framework (Draft) — Department of Communications and Digital Technologies | Superseded | Focuses on broad capabilities and scale warranting different regulatory treatment and higher scrutiny compared to narrower systems. |
What it triggers
- In Uzbekistan's Strategy for the Development of Artificial Intelligence Technologies until 2030, general-purpose AI is contrasted with narrow AI to inform long-term planning, risk consideration, and infrastructure development.source
- Under Romania's Draft Law B154/2024, general-purpose AI is described as a variant designed to perform tasks across multiple domains and contexts, with the potential to be adapted for diverse applications, subject to specific regulatory considerations.source
- South Africa's National Artificial Intelligence Policy Framework (Draft) describes general-purpose AI as warranting different regulatory treatment and higher scrutiny due to its scale and systemic effects.source
Definitions (4)
A variant of artificial intelligence designed or deployed to perform tasks across multiple domains and contexts rather than being narrowly specialized, with the potential to be adapted for diverse applications and thus subject to specific regulatory considerations.
Models characterized by broad capabilities across multiple tasks or domains (as distinct from narrow, application-specific systems). The Framework distinguishes these from narrower systems because their scale and systemic effects may warrant different regulatory treatment and higher scrutiny.
General-purpose AI describes more advanced AI systems intended to operate across multiple domains and perform a wide range of tasks with flexible reasoning and learning abilities. The strategy contrasts general-purpose AI with narrow AI for long-term planning, risk consideration, and infrastructure development.
General-purpose AI, referred to as foundation models, are large-scale models trained on broad datasets that can be adapted or fine-tuned for many different downstream applications. The Strategy highlights these models as a focal point for national research, governance, and risk-management given their broad applicability and systemic impact.
Not to be confused with
A general-purpose AI system refers specifically to an AI system based on a GPAI model serving multiple purposes, whereas general-purpose AI more broadly describes the foundational models or capabilities.
General-purpose AI (GPAI) model specifies a model trained at scale with broad task generality, while general-purpose AI encompasses systems or models performing across multiple domains.
GPAI is simply the abbreviation for general-purpose artificial intelligence, matching the term directly without adding distinct conceptual differences.
Questions and answers
How does Uzbekistan distinguish general-purpose AI from narrow AI?
General-purpose AI describes more advanced systems operating across multiple domains with flexible reasoning and learning abilities. The strategy contrasts it with narrow AI for purposes like long-term planning, risk consideration, and infrastructure development.
What regulatory considerations apply to general-purpose AI in Romania's draft law?
General-purpose AI is designed for tasks across multiple domains and contexts, with adaptation potential subjecting it to specific regulatory considerations like risk assessments and incident reporting.
Why does South Africa's policy framework treat general-purpose AI differently?
Models with broad capabilities across multiple tasks or domains are distinguished from narrow systems because their scale and systemic effects may warrant different regulatory treatment and higher scrutiny.
Related Terms
General-Purpose AI System
AI system based on GPAI model serving multiple purposes....
General-purpose AI (GPAI) model
Model trained at scale with broad task generality....
general-purpose technology
A technology with broad applicability across industries and society....
GPAI (General-Purpose AI)
Abbreviation for General-Purpose Artificial Intelligence....
General‑purpose AI model
AI models designed for multiple purposes, not a single specific task....