MAIAM v1 proposes a multi-tiered access and capability-limits framework for frontier AI
The paper (arXiv:2610.20921v1) proposes the Multi-Tiered AI Accessibility Model (MAIAM), which sorts access to AI by purpose, safety, and capability rather than by release artifact. It defines three tiers (broadly accessible, registered R&D, and tightly controlled full-scale access), and combines MAIAM with the SAFE AI Operating Point into a two-part framework—Capability Boundary (safety limits) and Accessibility Boundary (who may obtain capabilities)—while arguing that publishing frontier weights forfeits key safety levers and that safety fine-tuning can be removed from open weights at low cost.