Last updated:
The technical or capability-based limit (e.g., 10^25 FLOPs) that classifies a General-Purpose AI model as posing a high level of risk to society.
Under the EU AI Act, models exceeding this threshold face heightened obligations, including mandatory red-teaming and incident reporting to the EU AI Office. For enterprises, using a model that sits above this threshold means inheriting a supply chain that is under constant regulatory surveillance, which may impact model availability or performance.
Real world example:
A developer trains a new LLM using massive compute resources; because the training run exceeds 10^25 FLOPs, it is classified as a GPAI model with systemic risk, requiring a formal safety partnership with the EU.
