EU
← Back to Examples

Frontier AI Model with Systemic Risk

GPAI with Systemic Risk

A state-of-the-art general-purpose AI model trained with over 10^26 FLOPs of compute, capable of generating text, code, and reasoning across domains. Made available via API to millions of users worldwide.

Classification Walkthrough

  1. Step 1 — Prohibited practices (Article 5): The model does not perform any prohibited practice. ✗ Not prohibited.
  2. Step 2 — GPAI: The model is a general-purpose AI model. ✓ GPAI detected.
  3. Step 2a — Systemic risk: Training compute of 10^26 FLOPs exceeds the 10^25 threshold defined in Article 51. The model is automatically classified as posing systemic risk. ✓ Systemic risk.
  4. Result: GPAI WITH SYSTEMIC RISK — In addition to standard GPAI obligations (Articles 51-53), the provider must comply with Article 55: perform model evaluation including adversarial testing, assess and mitigate systemic risks, track and report serious incidents to the AI Office, and ensure adequate cybersecurity. Enforcement date: August 2, 2025.
Classify like thisView Checklist