In Washington, a new push for tiered AI safety rules aims to rein in giants without crushing startups

In Washington, D.C., the long‑running fight over how to regulate advanced “frontier” AI systems is heating up again. Recent reporting describes a bipartisan push for a tiered framework that would impose different obligations on tech giants and smaller players. The goal is to address mounting safety concerns without locking in today’s dominant platforms by saddling startups with the same compliance burden.

Under the emerging concept, obligations such as risk assessments, third‑party audits, and incident reporting would scale with factors like model compute, user reach, or deployment context. The most powerful systems would face stringent safety testing and external evaluations, while small firms and research labs would see lighter‑touch reporting rules.

Designing those tiers, however, is proving contentious. Lawmakers still need to define clear thresholds, figure out how to verify company claims, and reconcile clashing priorities: national security and economic competitiveness on one side, and aggressive safety guardrails on the other. Depending on how the eventual bill is written, global AI providers could face significantly different compliance requirements to access the U.S. market.