Dice Question Streamline Icon: https://streamlinehq.com

Danger from small‑scale deployment of frontier AI

Determine whether and under what conditions small‑scale deployment of frontier AI models—below the report’s large‑scale thresholds—can pose significant dangers, to inform whether verification should extend beyond large‑scale compute use.

Information Square Streamline Icon: https://streamlinehq.com

Background

The report focuses verification on large-scale compute use to balance feasibility and intrusiveness, but acknowledges a downside if dangerous deployment does not require scale.

Clarifying the risk profile of small‑scale frontier AI deployments would inform the appropriate scope and thresholds for verification regimes.

References

At the same time, this gap has downsides; it is not clear that frontier AI deployment must be done at scale to be dangerous.

— Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment (2507.15916 - Baker et al., 21 Jul 2025) in Section 2.2 (Rules on Large-Scale AI Compute)