Who Gets to Write the Rules for AI? Not You.

NewsOnScale Staff

October 22, 2026

There is a version of this story where democratic societies paused, deliberated, and collectively decided how one of the most consequential technologies in human history should be built and deployed. That is not the version unfolding right now.

Right now, the governance architecture for artificial intelligence — the rules determining how these systems can be used against you in a job application, a loan decision, a criminal sentencing recommendation, or a political advertisement — is being hammered out in legislative back-channels, agency comment periods that few know exist, and international standards bodies where corporate representation routinely outnumbers civil society by a ratio that would embarrass most democracies.

The public isn't in the room. And the people in the room are not rushing to open the door.

## A Legitimacy Problem Disguised as a Technical Problem

The standard defense of this process is that AI governance is complicated — that it requires technical expertise most citizens don't possess, that moving fast is necessary to keep pace with the technology itself, and that existing regulatory institutions are simply adapting to new circumstances. Each of these arguments contains a grain of truth. Together, they function as a permission structure for exclusion.

Complexity has never historically justified removing the public from decisions about technology that affects the public. Environmental regulation is complex. Drug approval is complex. Financial oversight is complex. In each of those domains, the United States and other democracies eventually built mechanisms — imperfect, contested, but real — for public accountability. Hearings. Comment periods with genuine reach. Advocacy organizations with enough standing and resources to push back against industry capture.

AI governance, by contrast, is moving at the speed of a closed negotiation. The technical standards being written today at bodies like the National Institute of Standards and Technology, or the interoperability frameworks being debated inside the EU's implementation of its AI Act, will shape what is legally permissible and commercially viable for decades. The window to influence those frameworks is not infinite. In many cases, it is already closing.

## What Exclusion Actually Costs

When governance processes are captured by the parties with the most at stake financially, the resulting rules tend to reflect those parties' risk tolerances — not society's. We have extensive documentation of what this looks like in practice: financial regulations written to accommodate the institutions they were meant to constrain, pharmaceutical approval pathways shaped by the companies seeking approval, social media content policies designed to protect engagement metrics rather than users.

AI governance conducted without meaningful public participation will produce the same dynamic, at greater scale and speed. High-risk applications — predictive policing tools, automated benefits determinations, hiring algorithms — will get frameworks calibrated to the operational preferences of the agencies and companies deploying them. The populations most likely to be harmed by those systems, who are also the populations with the least institutional access, will have had the least input.

This is not speculation. It is the observable pattern of technology governance conducted under time pressure with asymmetric stakeholder access.

## The Accountability Gap Is a Choice

It is worth being precise about what "the public isn't in the room" actually means mechanically. It does not mean that no public processes exist. It means that the processes that exist are often designed in ways — timing, technical opacity, resource requirements for meaningful participation — that systematically disadvantage non-industry voices.

Filling that gap requires deliberate structural choices: mandatory impact assessments with genuine public review periods, civil society representation with actual resources attached, and algorithmic transparency requirements that make it possible for journalists, researchers, and advocates to audit systems after deployment.

None of that is technologically difficult. All of it is politically difficult, precisely because the parties currently shaping these rules have limited incentive to share the table.

The decisions being made now will not wait for a more convenient political moment. Neither should the demand for accountability.