The Decisions That Will Define AI Are Being Made Without You

NewsOnScale Staff

October 24, 2026

There is a particular kind of democratic failure that doesn't look like a scandal. Nobody breaks a law. No documents get leaked. The process simply moves forward, technically open, functionally closed, and by the time ordinary people realize the rules have been written, the writing is done.

That is the pattern emerging in AI governance, and it deserves more scrutiny than it is currently receiving.

## Who Is Actually in the Room

Across the major forums where AI policy is taking shape — standards bodies, federal advisory committees, interagency working groups, and the informal convenings that precede formal rulemaking — the consistent presence is industry. The major AI developers, cloud infrastructure companies, and enterprise software vendors have the lobbyists, the subject-matter experts, and the scheduling flexibility to participate in months-long comment processes and working group sessions. Academics with the right affiliations get seats. Civil society organizations, when they are included at all, often lack the resources to engage at the depth and frequency required to shape outcomes rather than simply register objections.

This is not a conspiracy. It is a structural problem, and structural problems are often harder to fix than conspiracies because there is no single actor to hold accountable.

The result is that the people designing governance frameworks for AI systems are, in many cases, the same people building and selling those systems. The incentives do not have to be corrupt to be misaligned. They just have to be misaligned.

## What Is Being Decided Right Now

The stakes are not abstract. Active policy processes are determining whether AI systems used in hiring decisions will be subject to external auditing requirements or voluntary disclosure. They are determining what counts as a high-risk AI application and who decides when a system crosses that threshold. They are shaping whether algorithmic accountability will be enforced through existing civil rights law, new regulatory frameworks, or left primarily to market pressure and litigation.

Each of those questions has distributional consequences. The communities most likely to encounter AI in high-stakes contexts — in interactions with the benefits system, the criminal legal system, or automated hiring screens — are the least represented in the rooms where those questions are being answered.

## The Accountability Gap Is Not an Accident

Public participation in technical rulemaking has always been difficult. The processes are dense, the comment periods are short relative to the complexity of the subject matter, and the final rules rarely look like the summaries that reach general audiences. AI governance compounds these problems because the systems being governed are poorly understood even by many of their creators, because the harms are often probabilistic rather than immediate, and because the commercial interests opposing strong accountability measures have significant resources to deploy in advocacy.

None of that makes exclusion inevitable. It makes it a choice — one that policymakers and the institutions facilitating these processes are making, often by default, every time they design a working group without mandated public representation, or schedule a comment period without an accessible plain-language summary, or convene an expert roundtable whose experts are defined by proximity to capital.

## What Evidence-Based Accountability Looks Like Here

Covering this story responsibly means tracking who has a seat at specific tables, not just who is theoretically welcome. It means reading the actual membership rosters of federal advisory bodies, the participant lists from standards convenings, and the affiliation disclosures on submitted comments. It means asking not only what a proposed governance framework says, but who drafted the initial text and whose revisions survived.

It also means being honest about complexity. Not every industry participant has bad faith. Not every exclusion of the public is deliberate. Some of the people working inside these processes are genuinely trying to build frameworks that protect people. The problem is that good intentions operating inside a structurally captured process still produce captured outcomes.

The decisions being made right now about how AI systems will be governed are not preliminary. They are foundational. Frameworks established during this period will shape what is politically and legally possible for years afterward. The window for meaningful public influence is not indefinitely open.

The question worth asking — and asking persistently — is whether the people who will live with these rules had any real hand in writing them.