What happened
RAND's expert-insights perspective 'Settlement and Governance Behind the AI Frontier' argues that although 'much energy is devoted to debate about how to control frontier models, governance must extend to open-weight models and agents, which demand different conceptual frameworks and interventions.' The author contends that AI's power stems not just from machine intelligence but from the properties of the machines in which it is embedded — speed of dissemination, diversification, replicability, operational speed, extraordinary memory and low marginal cost — and that 'a distributed superintelligence is already here' as models and agents proliferate and coordinate, collectively achieving capabilities no single model possesses. It concludes that governing this ecosystem requires approaches different from regulating a handful of leading labs. Published 28 October 2026.
Why it matters
This challenges the frontier-model-centric framing that dominates AI governance, arguing that open-weight models and millions of agents already spread faster than regulators can track — directly relevant to anyone building AI governance or procurement policy.
Action needed
Re-examine your AI governance framework to ensure it covers open-weight models and agent ecosystems, not just frontier lab models.