AI's Rules Are Being Written Without You

NewsOnScale Staff

November 1, 2026

There is a particular kind of democratic failure that happens not through corruption or conspiracy, but through process — through the quiet accumulation of technical decisions made by people who were never elected, in meetings that were never announced, producing frameworks that will govern billions of people for decades. That is where artificial intelligence governance stands today.

The institutions shaping AI policy in 2026 — from the National Institute of Standards and Technology to the International Organization for Standardization to the Financial Stability Board — operate on timelines and in vocabularies that effectively exclude most citizens. Comment periods are short. Technical jargon is dense. Access to working groups often requires industry affiliation. And the stakes, almost universally, are framed as economic competitiveness rather than civil rights or democratic accountability.

This is not a neutral set of circumstances.

## Who Is Actually In the Room

The composition of AI governance bodies tells you a great deal about whose interests those bodies are likely to protect. Major technology companies maintain full-time government affairs teams dedicated to standards processes. Financial institutions have compliance departments that double as policy shops. Trade associations submit detailed technical comments that regulators, often understaffed and outgunned, struggle to meaningfully evaluate.

Civil society organizations, academic researchers, and public interest advocates participate where they can — but they are operating with a fraction of the resources, and they are often brought in at stages where fundamental architectural decisions have already been made. Consultation, in many of these processes, functions more as legitimation than as genuine input.

The result is governance frameworks that are fluent in the language of risk management and innovation incentives, and far less fluent in the language of worker displacement, algorithmic discrimination, surveillance, or the concentration of power.

## The Compounding Problem of Lock-In

What makes this moment particularly consequential is the degree to which early governance choices become self-reinforcing. Technical standards, once adopted, create infrastructure that is expensive to modify. Regulatory frameworks, once institutionalized, generate constituencies invested in their preservation. International agreements, once signed, constrain domestic policy space.

The companies that shape these early frameworks gain durable competitive advantages — not because their products are necessarily superior, but because the rules end up being written in ways that their existing systems satisfy and that new entrants struggle to meet. This dynamic, sometimes called regulatory capture, is not unique to AI, but the speed of AI development makes it unusually acute. The window between "rules being written" and "rules being locked in" is compressing.

## What Accountability Would Actually Look Like

Meaningful public participation in AI governance does not require that every citizen become an AI researcher. It requires that the people designing these systems be held accountable for explaining, in plain language, what they are building and who bears the costs when it fails. It requires that affected communities — workers whose jobs are being automated, people subject to algorithmic credit decisions, communities under AI-assisted surveillance — have genuine standing in the processes that govern those technologies, not just nominal access to comment boxes.

It also requires journalism, civil society, and legislative oversight that can match the complexity of what is being built. Legislatures in most democracies are still struggling to develop basic technical fluency. Investigative reporting on AI governance is thin relative to the scale of the decisions being made.

## The Honest Assessment

None of this means the people inside these governance processes are acting in bad faith. Many are genuinely trying to build responsible frameworks under significant time pressure. But good intentions do not substitute for structural accountability, and the structural accountability mechanisms for AI governance are, right now, badly underdeveloped.

The decisions being finalized in 2026 will shape what AI systems are permitted to do, who can challenge them, and who bears liability when they cause harm. Those decisions deserve a public that knows they are being made. Most people do not.