Why AI Policy Is Stuck: The Global Power Struggle Slowing the Rules That Matter

NewsOnScale Staff

July 6, 2026

There is no shortage of AI policy documents in the world. Roadmaps, white papers, governance frameworks, and interagency memos are being produced at a rate that rivals the technology they're meant to oversee. And yet, meaningful enforceable regulation — the kind with teeth, timelines, and accountability mechanisms — remains stubbornly out of reach in most of the world. The reason isn't technical complexity. It's political.

Governments are clashing over a deceptively simple question: who gets to set the rules? And until that question is answered, the rules themselves remain aspirational at best.

## A Three-Way Standoff

The conflict operates on at least three axes. First, there's the tension between national governments and supranational bodies like the European Union or UNESCO, each of which wants its governance model to become the global standard. Second, there's the friction between large AI-exporting nations — primarily the United States and China — who have competing economic and geopolitical incentives baked into whatever regulatory posture they adopt. Third, and perhaps most consequentially for everyday people, there's the standoff between governments and the technology industry itself, which has aggressively positioned voluntary self-governance as a preferable alternative to binding law.

The result is a regulatory environment that functions less like a framework and more like a negotiation that never quite closes. Proposals stall in committee. International coordination efforts produce declarations with no enforcement mechanisms. And in the vacuum, the systems being deployed — in hiring, credit, policing, healthcare, and civic infrastructure — operate largely on the terms of the companies that built them.

## Who Fills the Vacuum

This matters because governance vacuums don't stay empty. When binding regulation fails to materialize, informal power structures fill the space. Platform companies establish de facto standards through terms of service and model licenses. Industry consortia publish best-practice guidelines that are voluntary in name but treated as compliance proxies by regulators who have little else to point to. The optics of governance substitute for governance itself.

For the AI agent economy specifically, this dynamic is acutely dangerous. Autonomous systems making decisions on behalf of users — negotiating contracts, managing workflows, interacting with public services — require clear liability frameworks, audit rights, and disclosure requirements. None of those exist in coherent form in most jurisdictions. The question of who is legally responsible when an AI agent causes harm remains largely unanswered, not because it's unanswerable, but because no government has been willing to force the answer through the legislative process.

## The Sovereignty Problem

Underlying much of the gridlock is a sovereignty problem that rarely gets named directly. Nations that were slow to develop domestic AI capacity are now being asked to adopt governance frameworks written by, and largely favorable to, nations that weren't. That's not a paranoid reading — it's a structural reality that shapes how countries like Georgia, Brazil, or Kenya approach international AI governance talks. Accepting someone else's framework means accepting someone else's assumptions about acceptable risk, appropriate use, and who bears the burden of harm.

This is why copy-paste regulation — importing legal language wholesale from the EU AI Act or U.S. executive orders without adaptation — is not just technically insufficient. It's politically corrosive. It breeds distrust in the regulatory process itself, which in turn makes future cooperation harder.

## What Accountability Requires

None of this means regulation is impossible. It means it requires deliberate political will that has, so far, been consistently outmaneuvered by coordination failures and industry pressure. What accountability actually demands at this moment is less about having the perfect framework and more about being honest with the public about why the imperfect one keeps getting delayed.

The technology is not waiting. The deployment is not paused. Every month that governments spend relitigating jurisdiction is a month that consequential AI systems operate without meaningful oversight. That is a choice — and it belongs to the people making it.