Political Tech
As nations race to regulate artificial intelligence, competing interests and coordination failures are leaving the most consequential decisions unmade.
NewsOnScale Staff
July 5, 2026
There is a version of AI governance that works. Experts largely agree on its contours: clear liability standards, mandatory transparency for high-risk systems, meaningful public participation in rulemaking, and cross-border coordination strong enough to prevent regulatory arbitrage. Governments worldwide have had years to move toward that framework. Most haven't.
The gap between what AI governance requires and what it is actually delivering is now wide enough to drive a data center through.
Reports emerging from multiple policy institutions this week paint a consistent picture: the window for proactive, coherent regulation is closing, not because the technical problems are being solved, but because the political will to solve them is fragmenting. Governments that once seemed aligned on baseline principles are clashing over jurisdiction, competitive advantage, and control — while the systems they are supposed to be overseeing grow more capable by the quarter.
## The Coordination Problem Nobody Wants to Admit
The core difficulty isn't ideological. It's structural. AI systems don't respect national borders, but regulatory authority does. A model trained in one country, deployed through servers in a second, and used by citizens in a third produces a jurisdictional tangle that existing frameworks were never designed to handle.
The European Union has moved furthest with binding legislation, but its AI Act is already being tested by implementation gaps and industry pressure to soften enforcement timelines. The United States, under successive administrations, has swung between voluntary commitments and executive orders without passing comprehensive federal legislation. China has enacted sector-specific rules that prioritize state control over civil liberties. The result is a patchwork that sophisticated actors — large platforms, well-funded startups, state-backed enterprises — are already learning to navigate around.
International bodies have sounded the alarm. UN-affiliated panels have warned that unchecked AI deployment in critical infrastructure, financial systems, and information environments carries risks that are not merely theoretical. But warnings without enforcement mechanisms are just noise. And right now, the mechanisms don't exist at the scale the problem demands.
## Execution Is Where Governance Goes to Die
Even where laws exist on paper, the translation from policy to practice is faltering. Federal agencies in the United States, for example, have received direction to integrate AI governance into their operations — but without dedicated funding, technical expertise inside regulatory bodies, or clear accountability structures, those directives tend to dissolve into working groups and white papers.
This is not an accident. Regulatory capture is a well-documented phenomenon, and the AI industry has invested heavily in shaping the terms of its own oversight. The revolving door between major AI labs and government advisory roles is well-oiled. The think tanks that frame the dominant policy conversations are frequently funded by the companies those conversations are about. None of this is secret. Most of it is legal. All of it matters when evaluating who the resulting rules are actually designed to protect.
## What's Actually at Stake
The communities most exposed to AI's harms — those subject to automated hiring decisions, predictive policing systems, algorithmic benefits determinations, or AI-generated disinformation — are also the least represented in governance conversations dominated by industry lawyers and credentialed technocrats.
Location data offers a telling example. Precise geospatial information, increasingly processed by AI systems, enables surveillance at a scale and granularity that previous technologies never permitted. The regulatory frameworks governing its collection, sale, and use lag years behind the commercial reality. Brokers are selling it. Agencies are buying it. Rules that should constrain both are either absent or unenforced.
Governance that arrives after harm has already scaled isn't governance. It's documentation.
## The Accountability Gap Widens
What the current moment requires is not another framework document or stakeholder convening. It is legally binding, publicly enforceable rules with real penalties, developed through processes that include the people who bear the most risk — not just the people who stand to profit.
That kind of governance is technically achievable. The question of whether it is politically achievable is the one nobody in power seems eager to answer honestly.