Regulation (EU) 2024/1689

Article 51: Classification of general purpose AI models with systemic risk

Article 51 sets the line between an ordinary general purpose AI model and one with systemic risk. A model is presumed to carry systemic risk when the cumulative compute used for its training exceeds 10 to the power of 25 floating point operations, and the Commission can designate a model above or below that threshold.

Maximum penalty

Up to 15 million euro or 3 percent of worldwide annual turnover, whichever is higher.

Article 101

Binds from

2 August 2025

Obligations applied from 2 August 2025. The Commission's power to fine for breach of them, under Article 101, applied from 2 August 2026.

Enforced by

The European Commission through the AI Office, with exclusive competence under Article 88.

Plain reading

What Article 51 actually requires

The classification decides how heavy the obligations get. Every general purpose model provider owes the Article 53 duties: technical documentation, information for downstream providers, a copyright policy and a public summary of training content. A provider whose model is classified as carrying systemic risk owes Article 55 on top of that: model evaluation including adversarial testing, systemic risk assessment and mitigation, serious incident reporting to the AI Office, and adequate cybersecurity. Providers must notify the Commission without delay, and in any event within two weeks, when their model meets or is expected to meet the threshold. A provider may argue that its model does not present systemic risk despite crossing the compute threshold, but it must make that argument to the Commission rather than decide unilaterally.

The compute threshold is a presumption, not a definition. Article 51(1)(a) also allows designation on the basis of capabilities or impact equivalent to the models above the threshold, on the criteria in Annex XIII, which means a highly capable smaller model is reachable.

The Commission may amend the threshold by delegated act to keep pace with the state of the art. Any provider treating 10 to the power of 25 as a permanent safe harbour is planning against a number the Act was drafted to allow the Commission to move.

Notification runs to the Commission, not to a national authority. General purpose models are the one part of the Act where the Commission, through the AI Office, holds exclusive enforcement competence under Article 88.

No model has been the subject of a public systemic risk enforcement action. The AI Office's stated opening posture is informal technical compliance dialogue rather than formal proceedings.

The official text, consolidated with the Digital Omnibus applied, with everything that cites it and everything it cites: Article 51 in the EU AI Act Explorer.

Enforcement

Actions citing Article 51

No tracked enforcement action expressly cites this article. The list below is the conduct that falls inside it, enforced so far under the GDPR or the Digital Services Act instead.

No enforcement action on this tracker touches Article 51 yet. That is the accurate answer as of 20 August 2026, and it is checked every Monday. When the first one lands it appears here the same day.

Answers

Questions about Article 51

What makes a general purpose AI model systemic risk under the EU AI Act?

Article 51 presumes systemic risk where the cumulative compute used for training exceeds 10 to the power of 25 floating point operations. The Commission may also designate a model on the basis of equivalent capabilities or impact using the Annex XIII criteria, and may amend the threshold by delegated act.

The rest of the Act

Other enforceable provisions

Cite this page

Free to reuse under CC BY 4.0, with attribution.

In a sentence: According to the GAGE EU AI Act Enforcement Tracker (as of 20 August 2026), eu ai act article 51: classification of general purpose ai models with systemic risk.
APA: GAGE (Global Academy of Generative-AI Education). (2026). EU AI Act Article 51: Classification of general purpose AI models with systemic risk. EU AI Act Enforcement Tracker. Retrieved 20 August 2026, from https://www.gage.academy/tools/eu-ai-act-enforcement-tracker/provisions/article-51
MLA: "EU AI Act Article 51: Classification of general purpose AI models with systemic risk." EU AI Act Enforcement Tracker, GAGE (Global Academy of Generative-AI Education), 20 August 2026, https://www.gage.academy/tools/eu-ai-act-enforcement-tracker/provisions/article-51.
Chicago: GAGE (Global Academy of Generative-AI Education). "EU AI Act Article 51: Classification of general purpose AI models with systemic risk." EU AI Act Enforcement Tracker. Last modified 20 August 2026. https://www.gage.academy/tools/eu-ai-act-enforcement-tracker/provisions/article-51.
Permalink: https://www.gage.academy/tools/eu-ai-act-enforcement-tracker/provisions/article-51

Last updated 20 August 2026. Every record re verified 20 August 2026. The tracker is checked every Monday, and the same day for any action under the AI Act.