GPAI: General-Purpose AI as Regulatory Category
The EU AI Act introduced “General-Purpose AI” (GPAI) as a distinct regulatory category, responding to the emergence of foundation models like GPT.
GPAI is defined as an AI model that “displays significant generality and is capable of competently performing a wide range of distinct tasks regardless of the way the model is placed on the market.” This contrasts with traditional narrow AI systems designed for specific applications.
The GPAI designation creates a parallel governance track separate from the Act’s application-based risk classification. When a GPAI model is integrated into a system that can serve multiple purposes, the entire system becomes a “GPAI system” with distinct compliance requirements.
GPAI models with “systemic risk” face additional obligations. The Act uses a threshold of 10^25 FLOPs of cumulative compute for training, models above this must comply with enhanced requirements including model evaluations, adversarial testing, and incident tracking.