OpenAI backs US ‘reverse federalism’ path to a national AI safety standard

This is the clearest articulation yet of OpenAI’s preferred regulatory architecture: let a handful of big blue states pass aligned frontier‑safety laws, then use that momentum to justify a single federal regime that pre‑empts a patchwork and becomes the template for a US‑led global framework. Lehane’s piece reads less like a blog and more like a bid to define the Overton window: national security‑grade testing run by Washington, mandatory incident reporting and audits for frontier labs, and states nudged away from ad‑hoc restrictions toward a harmonised baseline.([openai.com](https://openai.com/index/advancing-ai-safety-through-state-and-federal-action/?utm_source=openai))

For the AGI race, this is effectively OpenAI arguing for “soft nationalization” of the safety gate while keeping the labs private. If the federal government controls who gets access to the highest‑risk models and how they’re tested, then frontier capability starts to look like a regulated utility. That could entrench today’s top labs – they’re the ones with the resources to comply – while making it harder for upstarts to compete at the very top end of capability.

The flip side is that a coherent national framework may also accelerate deployment to critical infrastructure, allies and defence, because there’s finally a playbook for getting them privileged access. If that happens, the US doesn’t just lead on raw model capability; it leads on how those models are embedded into the security and economic apparatus – a different, but equally important, axis in the race to AGI.