Political Tech
As governments and industry groups race to write the foundational rules for artificial intelligence, the democratic deficit at the heart of that process is growing harder to ignore.
NewsOnScale Staff
October 25, 2026
There is a version of the AI governance story that gets told often: a race between the United States and China, a contest of regulatory philosophies, a question of whether innovation will be strangled by red tape or weaponized by bad actors. It is a real story. It is also a story that consistently centers the wrong people.
The more uncomfortable question — the one that tends to get edited out of conference agendas and whitepapers — is who exactly is in the room when these frameworks get written, and who is systematically not.
The answer, increasingly documented by researchers and civil society organizations tracking the governance process, is that the rooms are full of industry representatives, platform executives, and government officials, and nearly empty of the workers whose labor trains the models, the communities whose data fuels them, and the citizens who will encounter AI-driven decisions in their housing applications, parole hearings, and medical diagnoses.
## A Legitimacy Problem Dressed as a Technical One
Governance processes routinely frame public exclusion as a practical necessity. AI systems are complex, the argument goes. You need technical expertise in the room. Comment periods exist. There are advisory boards.
But advisory boards are not binding. Comment periods are frequently pro forma. And the definition of who counts as a technical expert has a way of defaulting to people employed by the companies being regulated.
This is not a conspiracy — it is a structural pattern. Regulatory capture does not require bad intent. It requires only that the parties with the most resources and the most to gain from particular rule-sets are also the parties with the most consistent, most organized access to the people writing those rules. Over time, the Overton window for what governance can plausibly look like narrows to what the best-resourced voices in the room consider acceptable.
The result is frameworks that speak fluently about innovation corridors and liability limitations, and more vaguely about enforcement mechanisms, community redress, or mandatory transparency requirements with actual teeth.
## What Accountability Requires
Accountability-focused governance — the kind that holds up over time — requires a few things that current processes largely lack.
First, it requires meaningful representation from affected communities at the drafting stage, not the comment stage. There is a significant difference between shaping a rule and responding to one that has already been written. Organizations working on AI policy from a civil rights or labor perspective have noted repeatedly that they are invited to react, rarely to co-construct.
Second, it requires transparency about who is in the room. Many AI governance working groups and advisory committees do not publish their membership, their deliberations, or the specific provisions that were proposed and rejected. Without that record, public accountability is nearly impossible. You cannot audit a process you cannot see.
Third, it requires enforcement infrastructure that does not assume good faith compliance. Rules that rely primarily on voluntary adoption or self-certification reproduce the same problem in a new form. The history of platform self-regulation over the past two decades should have settled this question by now.
## The Window Is Closing
Governance frameworks, once embedded, are extremely difficult to revise. The technical standards, liability structures, and institutional habits that get built into first-generation AI rules tend to persist. Incumbents have every incentive to lock in frameworks that favor their current market position. That is not cynicism — it is how regulatory ecosystems work.
This means the democratic deficit in today's AI governance process is not just a procedural concern. It is a compounding one. Every month that foundational rules get written without genuine public representation is a month that makes subsequent course correction harder and more contested.
The race worth paying attention to is not the one between Washington and Beijing. It is the one between the speed of rule-writing and the speed at which affected communities can organize enough to matter. Right now, the rules are winning.