What is a GPAI Model?
Under Article 3(63) of the EU AI Act, a General Purpose AI (GPAI) model is an AI model trained on large amounts of data, designed to serve a variety of purposes, and which can be integrated into downstream systems or applications.
This explicitly includes: GPT-4, Claude, Gemini, Llama, Mistral, and any other large language model or multimodal foundation model used as a base layer in a product.
Who is Affected — Provider vs Deployer
GPAI Model Providers (OpenAI, Anthropic, Google...)
Subject to the full Chapter V obligations — technical documentation, copyright policy, transparency to downstream providers.
Downstream Deployers (you, if you build on a GPAI model)
Subject to Article 50 transparency obligations and potentially high-risk obligations if your use case falls under Annex III.
Article 53 — Obligations for GPAI Providers
Article 55 — Systemic Risk Models
GPAI models trained with more than 10²⁵ FLOPs are classified as systemic risk models (currently: GPT-4, Gemini Ultra, Claude 3 Opus, Llama 3 405B). Additional obligations apply:
What Deployers Must Do in Practice
If you are building a product on a GPAI model, your obligations depend on your use case:
If your use case is NOT in Annex III
Article 50 applies — disclose to users that they are interacting with an AI system. For chatbots: clear disclosure at the start of each interaction.
If your use case IS in Annex III (HR, credit, healthcare...)
The full high-risk regime applies to your system — even if the underlying model is not itself high-risk. You are the provider of a high-risk AI system.
Timeline
Using a GPAI model in your product? The free Sprinkling Act diagnostic determines whether Article 50, Article 53, or the full high-risk regime applies to your system.
Sources
- [1]EUR-Lex (July 12, 2024) — Regulation (EU) 2024/1689 — Artificial Intelligence Act (full text) eur-lex.europa.eu/eli
- [2]
- [3]EU AI Act — Article 51 — Classification of General-Purpose AI Models as GPAI Models with Systemic Risk artificialintelligenceact.eu/article
- [4]EU AI Act — Article 53 — Obligations for Providers of General-Purpose AI Models artificialintelligenceact.eu/article
- [5]EU AI Act — Article 55 — Obligations for Providers of GPAI Models with Systemic Risk artificialintelligenceact.eu/article
- [6]EU AI Act — Article 50 — Transparency Obligations for Deployers artificialintelligenceact.eu/article
- [7]EU AI Act — Annex III — High-Risk AI Systems Referred to in Article 6(2) artificialintelligenceact.eu/annex
- [8]
Les interdictions Art. 5 et les règles GPAI s’appliquent aujourd’hui. La transparence suit dans 105 jours. La question n’est pas quand — c’est si vous avez documenté votre position.