Biden's AI Executive Order Is Back in the Conversation, and Its Defenders Are Being Specific

Someone on Bluesky this week pointed at the wreckage of the current AI regulation landscape and said the early contours of what a real policy response should look like were visible in Biden's AI Executive Order[¹] — and the observation landed harder than it probably would have six months ago. The post got 44 likes in a community that doesn't typically reward nostalgia. What made it stick wasn't sentiment. It was specificity: the commenter wasn't arguing that the EO was visionary or sufficient, but that it represented a kind of sector-appropriate, layered governance logic that the current moment conspicuously lacks. That framing — not "Biden was right" but "we at least had a map" — is doing something interesting in a conversation that has otherwise bottomed out into bans-versus-nothing.

The argument arrived at a specific moment of regulatory disarray. Maine's governor recently vetoed what would have been the country's first statewide data center moratorium, and the global proliferation of AI laws has done little to clarify what enforcement actually looks like. Meanwhile, the White House has been signaling procurement guidelines as a substitute for binding rules — governance by vendor relationship rather than statute. Against that backdrop, a separate voice in the same thread made the more pointed observation: the communities most energized about AI policy right now are debating data center bans, not model governance, not labor protections, not sector-specific liability. The question of where to locate a server farm has become the proxy war for a much larger argument that nobody has figured out how to have yet. One commenter put it plainly: bans don't slow AI development, they just relocate it — and regulation, however imperfect, is the only intervention that actually touches the system[²].

The data center thread is a useful lens because it reveals the gap between what regulators can see and what they need to govern. A moratorium on physical infrastructure is legible, mappable, and enforceable in ways that, say, model transparency requirements are not. That's partly why it attracts legislative energy — and partly why it frustrates people who think it's the wrong target entirely. The compute argument is real: you can't run cloud AI without local infrastructure, and concentrated compute capacity is a genuine chokepoint. But the communities most invested in AI governance — the ones building frameworks, writing logging standards, pushing for sector-appropriate rules — are mostly not talking about data centers. They're talking about the harder, less photogenic work of what enterprise governance actually requires, and whether any institution is structured to demand it.

What the Biden EO nostalgia actually represents isn't a policy prescription — it's a measurement tool. The people invoking it aren't proposing its reinstatement; they're using it to show the distance between where the conversation was and where it has drifted. That distance is the story. When the most coherent federal AI policy in recent memory becomes a reference point for what's missing rather than what was achieved, the governance conversation hasn't moved forward. It's moved sideways into infrastructure fights and preemption battles while the harder questions — liability, transparency, sector risk — accumulate without a home.