The Room Where It Happens: AI's Rules Are Being Written Without You

NewsOnScale Staff

October 23, 2026

There is a version of democratic governance where citizens shape the rules that govern powerful technologies before those technologies reshape their lives. That version is not what is happening right now with artificial intelligence.

Across Washington, Brussels, Beijing, and Geneva, the scaffolding of AI governance is being erected at speed — in technical standards bodies, regulatory working groups, bilateral diplomatic summits, and industry-led forums where participation requires either a badge or a wire transfer. The decisions being locked in now will determine how AI systems are audited, who bears liability when they cause harm, what data can be used to train them, and which countries get to export their legal frameworks as de facto global standards. The public, by nearly every available measure, is not meaningfully present for any of it.

## Speed as a Substitution for Legitimacy

The urgency argument is familiar by now: AI is moving fast, governance must keep pace, there is no time for slow deliberation. It is not an entirely dishonest argument. The gap between AI capability and regulatory capacity is real, and the costs of that gap — in labor displacement, algorithmic discrimination, national security exposure, and democratic manipulation — are accumulating without a corresponding accountability structure to absorb them.

But speed has a politics. When governance moves faster than public engagement can realistically track, the stakeholders who were already in the room before the sprint began are the ones who shape the finish line. Those stakeholders are, predictably, the large technology companies, well-resourced industry associations, and government actors who have the institutional bandwidth to engage with complex technical rulemaking processes in real time.

This is not a conspiracy. It is a structural feature of how regulatory capture happens in technically complex domains — not through bribery or bad faith, but through the simple arithmetic of who has the time, expertise, and access to participate continuously while everyone else is busy.

## What Is Actually Being Decided

The stakes deserve specificity. Among the questions currently being resolved in forums with limited public accountability: whether AI systems used in hiring, lending, and criminal justice should face mandatory third-party audits, and if so, who accredits the auditors. How liability should be allocated when an AI system causes harm — between the developer, the deployer, and the end user. What transparency obligations should attach to foundation models. Whether national security carve-outs will effectively immunize entire categories of high-risk AI from civilian oversight.

These are not merely technical questions dressed up in policy language. They are distributive questions about power — about who can be harmed, who must account for that harm, and who gets to decide what counts as harm in the first place.

## The Participation Infrastructure Doesn't Exist Yet

Criticizing the absence of public participation is easy. Describing what genuine participation would require is harder and more useful. It would require, at minimum: plain-language summaries of proposals published before comment periods close rather than after; funded civil society representation in technical standards processes that currently charge participation fees; mandatory disclosure of which industry representatives participated in drafting regulatory text; and impact assessments that evaluate governance proposals against the interests of communities most likely to bear the costs of AI deployment.

None of that infrastructure exists at meaningful scale. Some of it is being built — the AI Safety Institute model in the UK, the EU AI Act's post-market monitoring requirements, various multi-stakeholder initiatives with real but uneven records. But building participation infrastructure after the foundational rules are already set is a different exercise than building it before.

## The Window Is Narrow

Governance frameworks, once adopted, develop institutional inertia. The compliance ecosystems, legal interpretations, and diplomatic precedents that accumulate around early frameworks make them resistant to revision even when their flaws become apparent. The EU's GDPR, for all its ambition, illustrates this dynamic: a framework genuinely aimed at privacy protection that has been operationally colonized by consent-theater that serves no one except the lawyers who bill to maintain it.

The same fate is available for AI governance. The question is whether the window for meaningful course correction — which requires knowing what is being decided and by whom — stays open long enough for anyone outside the room to walk through it.