Art.52

Article 52 — Procedure for Classifying GPAI Models with Systemic Risk

GPAI Models SME Relevant ~2 min read · 457 words

Article 52 sets out the procedural rules for classifying GPAI models as presenting systemic risk — including the notification obligation, the process for providers to challenge a classification, and the Commission’s investigative powers. It is the procedural engine behind the classification criteria of Article 51.

✓ Low compliance impact for SMEs

WHAT THE ARTICLE IS ABOUT

How systemic risk classification is initiated, challenged and confirmed

Article 52 establishes the procedure by which GPAI models are formally classified as presenting systemic risk under Article 51. It covers notification obligations, the Commission’s investigative powers, the provider’s right to present arguments against classification, and how the scientific panel of independent experts feeds into the process.

WHAT IT SAYS

Notify within two weeks, challenge within time limits, Commission decides

Providers of GPAI models that meet or approach the 10^25 FLOPs threshold must notify the Commission within two weeks of the model meeting the threshold
The Commission may request additional information from providers to assess whether a model presents systemic risk — providers must respond within a specified deadline
Where the Commission has reason to believe a model presents systemic risk — whether or not it meets the compute threshold — it must carry out an assessment, which may include requesting technical documentation, model access, or expert input
The scientific panel of independent experts established under Article 68 may submit qualified alerts to the Commission identifying models that may present systemic risk
Providers may submit arguments to the Commission demonstrating that their model, despite meeting the threshold, does not present systemic risk — the Commission must consider these arguments before confirming classification
The Commission’s decisions classifying a model as systemic risk are published and may be appealed through standard EU administrative review procedures

WHO IS AFFECTED

Large foundation model providers and the Commission’s AI Office

Providers of frontier AI models approaching or exceeding the compute threshold
The European Commission’s AI Office which manages the classification procedure
The scientific panel of independent AI experts who may trigger classification proceedings
Other member states and stakeholders who may be consulted during classification assessments

WHAT IT MEANS FOR SMES

Background governance — but shapes the AI models you depend on

This procedural article has no direct obligations for SMEs — it governs the relationship between frontier AI developers and the Commission
Indirectly, it matters because the classification procedure determines which AI models face the heaviest regulatory scrutiny — and therefore which models have the strongest compliance frameworks underpinning them
The two-week notification requirement creates near-real-time Commission visibility into the most powerful AI models being trained in or deployed in the EU
The challenge procedure is important: even if a model meets the compute threshold, its provider can argue it does not present systemic risk — this prevents automatic classification from creating disproportionate compliance burdens for niche high-compute models that pose limited actual risk

Article 51 — Classification criteria (the threshold this procedure operationalises)
Article 55 — Systemic risk obligations (what classification triggers for the provider)
Article 68 — Scientific panel of independent experts (the body that can trigger classification via qualified alerts)
Article 88 — Enforcement of GPAI obligations (the enforcement framework for systemic-risk providers)

← Back to the Guide