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

NewsOnScale Staff

November 3, 2026

There is a particular kind of democratic failure that doesn't look like a failure at all. No laws are broken. No votes are suppressed. The meetings happen, the documents get published, and the press releases go out. But somewhere in the machinery of process, the public — the people who will actually live inside the systems being designed — gets quietly cut out of the room.

That is what is happening right now with AI governance, and it is happening fast.

## The Illusion of an Open Process

Most people assume that rules with the force of law go through something resembling democratic deliberation. For AI, that assumption is dangerously out of date. The frameworks being constructed today — the ones that will determine how AI systems make decisions about credit, hiring, healthcare triage, criminal risk scores, and content moderation at scale — are emerging from a patchwork of industry working groups, agency guidance documents, and international standards bodies where corporate representation vastly outnumbers civil society.

This is not a conspiracy. It is the predictable outcome of a system where technical complexity is used, intentionally or not, as a barrier to entry. When the conversation requires fluency in model cards, algorithmic auditing standards, and the difference between ex-ante and ex-post regulatory approaches, the organizations with armies of policy staff and lobbyists have a structural advantage that no amount of public comment periods can fully offset.

The result is that governance frameworks get written to reflect the risk tolerance, liability preferences, and competitive positioning of the entities that showed up — not the populations most exposed to harm.

## What's Actually Being Decided

The stakes are not abstract. The technical standards being negotiated now will determine whether an AI system used to evaluate a job application must explain its reasoning, or whether 'explainability' is defined so narrowly that a one-line output counts as compliance. They will determine whether a company deploying an AI model in a high-risk context bears meaningful liability when it causes harm, or whether that liability gets diffused across a supply chain of vendors until it effectively disappears.

They will also determine who gets to audit these systems — whether independent researchers and regulators have genuine access to model behavior, training data, and performance metrics, or whether 'transparency' means a company publishes a glossy responsibility report once a year.

These are policy choices dressed in technical language. And they are being made on a timeline driven by market competition, not democratic readiness.

## The Preemption Problem

One of the least-discussed dynamics in AI governance is what happens at the state and local level when federal frameworks — or voluntary industry standards — effectively crowd out more protective rules before they can be enacted. When a federal agency issues guidance or when a major standards body ratifies a framework, it creates political and legal pressure on state legislators to conform. More aggressive consumer protections, algorithmic accountability requirements, or sector-specific rules can become legally or politically untenable almost overnight.

This dynamic rewards the organizations that can shape frameworks early, at the top of the hierarchy, before any democratic constituency has organized around an alternative.

## What Accountability Would Actually Look Like

Public participation in AI governance doesn't have to mean turning every technical standard into a ballot initiative. But it does mean something more than a 60-day comment window that regulators have no structural obligation to act on. It means civil society organizations with resources to match industry in technical depth. It means mandatory disclosure of who was at the table when draft frameworks were written. It means transparency about which provisions changed between the first draft and the final rule, and at whose request.

It means treating AI governance less like a technical compliance exercise and more like what it actually is: a set of political decisions about power, accountability, and who bears risk when systems fail.

The window is not closed yet. But the concrete is being poured. And once the foundational rules are set, reforming them will require fighting not just inertia, but the entrenched interests of every organization that built a business model around the current architecture.

That is worth paying attention to now — before the room empties and the door locks behind it.