General-purpose AI model (GPAI)
Models adaptable to multiple tasks or domains.
How the laws define this
Regulators consistently define a general-purpose AI model as one designed or intended to perform a broad range or spectrum of tasks across multiple contexts, domains, or downstream systems, rather than being limited to a single, narrowly defined purpose. Standard requirements across these frameworks include provider obligations regarding transparency, technical documentation, and systemic risk mitigation. However, these instruments diverge in how they specify the scope and trigger criteria for regulatory oversight. The European Union Regulation on Artificial Intelligence applicable in Belgium and Ireland emphasizes design, deployment, or marketing to impose provider duties like safety assessments, watermarking, and cooperation with the European Artificial Intelligence Office. Conversely, the Proposal for a Digital Omnibus on Artificial Intelligence Amending Regulation (EU) 2024/1689 anchors its definition in specific technical capabilities, requiring models to display significant generality and competent performance through self-supervision at scale using large data amounts. Meanwhile, EDPB-EDPS Joint Opinion 1/2026 focuses on a model's usability for numerous downstream systems to centralise oversight.
Synthesised from the 6 statutory definitions below. Each one is quoted in full, with its source.
What it triggers
- Under Regulation (EU) 2026/1744 of the European Parliament and of the Council of 8 July 2026 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI), GPAI models are subject to simplified AI compliance obligations for developers and deployers.source
- Under the European Union Regulation on Artificial Intelligence (AI Act) — as applicable in the Czech Republic, GPAI models are subject to special transparency, documentation and governance obligations, including requirements for provider disclosures and risk mitigation when outputs are used in the Union.
- Under Regulation (EU) 2024/1689 — Artificial Intelligence Act (EU AI Act) (applicable in Ireland as an EU Regulation), GPAI models are subject to Chapter V obligations including systemic risk identification and mitigation, tailored technical documentation, transparency duties and cooperation with the European Artificial Intelligence Office.
- Under Regulation (EU) 2024/1689 — Artificial Intelligence Act, GPAI models are subject to specific documentation, testing, watermarking/metadata and mitigation measures to address scale and systemic risks, with the Commission empowered to set further obligations and fines for non-compliance.
Definitions (6)
A model designed to perform a broad range of tasks and generate outputs across multiple contexts rather than being limited to a specific narrowly defined purpose; the Act specifies special transparency, documentation and governance obligations for GPAI, including requirements for provider disclosures and risk mitigation when outputs are used in the Union.
A model designed or intended to perform a broad spectrum of tasks rather than being tailored to a specific application; GPAI are subject to Chapter V obligations including systemic risk identification and mitigation, tailored technical documentation, transparency duties and cooperation with the European Artificial Intelligence Office.
AI models designed and marketed for general-purpose use across diverse tasks and contexts rather than a specific narrowly defined purpose; the Act imposes specific documentation, testing, watermarking/metadata and mitigation measures on GPAI providers to address scale and systemic risks, and empowers the Commission to set further obligations and fines for non-compliance.
AI models designed and deployed to perform a wide range of tasks across different contexts and domains rather than a narrowly defined specific purpose; the Act defines GPAI and subjects their providers to specific transparency, documentation, safety assessment and Commission oversight obligations, including methodologies for identifying systemic risk. GPAI may attract additional regulatory scrutiny due to potential systemic effects and broad deployability.
An AI model not designed for a single narrow purpose but usable for numerous downstream systems; the Digital Omnibus proposal centralises oversight of many systems built on certain GPAI categories, requiring clear delimitation of types triggering AI Office competence.
GPAI models are defined as AI models displaying significant generality and capable of competently performing a wide range of distinct tasks, often trained with large amounts of data using self-supervision at scale.
Not to be confused with
Model for general-purpose use refers to an AI model intended for broad, non-domain-limited tasks, while GPAI emphasizes significant generality, competence across a wide range of distinct tasks, and specific regulatory obligations.
Model AI of General Purpose describes a model designed to perform broad tasks across domains, whereas GPAI additionally highlights self-supervision at scale, systemic risk obligations, and broad deployability across contexts.
Questions and answers
What obligations apply to GPAI models under the EU AI Act?
GPAI models are subject to specific transparency, documentation, safety assessment, and risk mitigation obligations. Providers must comply with Chapter V requirements including systemic risk identification, technical documentation, and cooperation with the European Artificial Intelligence Office. Additional measures like watermarking and metadata apply to address systemic risks.
How does the Digital Omnibus on AI affect GPAI models?
Regulation (EU) 2026/1744 simplifies AI compliance obligations for developers and deployers of GPAI models. It centralises oversight for systems built on certain GPAI categories and requires clear delimitation of types triggering AI Office competence.
What distinguishes GPAI from narrow AI models?
GPAI models are designed for a broad range of tasks across multiple contexts and domains, not limited to a specific narrowly defined purpose. They display significant generality and are capable of competently performing diverse tasks, often trained with large amounts of data using self-supervision at scale.
Related Terms
General‑Purpose AI Model (GPAI)
Models designed for broad use across tasks and contexts....
General‑purpose AI model
AI models designed for multiple purposes, not a single specific task....
Model for general-purpose use
AI model intended for broad, non-domain-limited tasks....
General‑Purpose AI (GPAI) model
AI systems intended to perform diverse tasks and be adapted for downstream uses....
Model AI of General Purpose
A model designed to perform broad tasks across domains....