Article 51 — Classification of general-purpose AI models as general-purpose AI models with systemic risk

A general-purpose AI model counts as carrying systemic risk when it has high-impact capabilities, judged through technical benchmarks and indicators. Training compute above 10^25 floating-point operations creates a presumption of such capabilities. The Commission may also designate a model on its own initiative or after a qualified scientific-panel alert, using the criteria in Annex XIII, and can update the thresholds and benchmarks as the technology advances.