The world trusts China to regulate AI more than it trusts America

Pew Research Center's 37-country survey, published today, contains a number the AI industry has spent two years treating as settled: when people in 12 countries were asked which government they trust to regulate artificial intelligence effectively, China came first. A year ago the same question put the European Union on top, the United States second and China last. That ordering is now reversed, and it arrives in the same week Washington finished dismantling its own frontier oversight.

The inversion nobody priced

The margin is not close. Across the 12 countries Pew asked — 13,696 adults surveyed between February 8 and May 13 this year — the median who trust China to regulate AI effectively is 43%, against 35% for the United States and 34% for the European Union. China is the most-trusted regulator in Bangladesh, Malaysia, Pakistan, Sri Lanka and the West Bank and East Jerusalem; Singaporeans split between China and the EU; the Philippines is the only country surveyed where the US leads both.

Pew's own read is that this is not a considered verdict on regulatory machinery. Trust tracks overall favorability almost mechanically: among Peruvians who view China favorably, 57% trust it to regulate AI, against 26% of those who view it unfavorably. So the honest interpretation is narrower and more useful — the question is measuring which power middle-income countries are geopolitically aligned with, not which one writes better rules. In a year, the alignment has moved.

The domestically focused numbers are just as uncomfortable for Washington. In half the countries surveyed, people trust their own government more than any of the three, and that faith is concentrated where it should be least reassuring: about eight in ten adults in Pakistan and Bangladesh trust their governments to regulate AI effectively. The US has the opposite problem. Pew's parallel American polling this summer found young adults' concern climbing sharply, which is a political liability no framework has yet answered.

The jobs number is a wealth gradient. The fear isn't.

In 34 of 37 countries, people expect AI to mean fewer jobs over the next 20 years rather than more. That expectation splits cleanly along income: a median of 55% in 18 high-income countries expect fewer jobs, against 36% across 18 middle-income countries — where 34% simply say they are not sure, against 22% in rich countries.

The inequality question follows gross domestic product per capita almost exactly. A median of 35% in high-income countries say AI will widen the gap between rich and poor, against 22% in middle-income nations: 46% in the US, 13% in Colombia. Concern about AI's role in daily life is far less GDP-linked — 40% of high-income respondents are primarily concerned against 31% in middle-income countries, while excitement sits at 13–14% everywhere and a 41% global median reports being both concerned and excited. The single most useful number in the report may be the awareness gap: half of adults in high-income countries have heard or read a lot about AI, against 27% in middle-income ones, with 28% of the latter saying they have heard nothing at all.

Where the gradient does break is age, and it breaks toward pessimism. In all 36 countries with comparable data, adults aged 18 to 34 are more likely than those 50 and older to have heard a lot about AI — by 51 percentage points in Spain. More importantly, their concern has converged with the oldest cohort's. In the US, the share of 18- to 34-year-olds more concerned than excited jumped from 40% in 2024 to 55% today, while older cohorts held flat. Sweden flipped outright: last year 24% of under-35s were primarily concerned against 36% of over-50s; now it is 45% and 43%. The people entering the labor market are now as worried as the people leaving it — a turn that started in the US and is now showing up in Sweden, Poland, Brazil, Japan and Australia.

Washington vacated the role. Beijing staffed it.

Trust numbers matter because rules are contested territory, and the two venues that could set them have moved in opposite directions this year. The White House drafted a frontier oversight body this summer and shelved it — reportedly after Mark Zuckerberg, Jensen Huang and Elon Musk made the case against regulation directly to the president — leaving the Frontier Act and the kill-switch bills with no floor time before the midterms. The full shape of that vacuum is in The AI oversight body died in draft — and nothing replaces it: the only functioning US mechanism is voluntary disclosure, and the lab writes the template.

China, meanwhile, has spent 2026 shipping rulebooks rather than debating them: labels and human sign-off required on every AI-made broadcast program, runaway agents folded into its TC260 safety framework, and AI standards used as an export channel to smaller markets through APEC. If you want the background on how different the two systems' instruments are, AI 101 — What is AI regulation? walks through what each jurisdiction's rules actually cover.

The skeptic's case against reading too much into the trust number is strong, and worth stating plainly. Trusting a government to regulate a technology is not evidence that its regulation works — Beijing's AI rules are substantially content controls, enforced by the same apparatus that censors speech, and none of them has been shown to make a model safer. A single survey question asked in 12 mostly middle-income countries is thin evidence of a global realignment. And the fieldwork closed on May 13, before the Hugging Face agent breakout, before the pacing essays, before this month's warnings — the survey measures a mood that the summer may already have rewritten. Pew says as much by dating it.

But the trend line is the point, and it is not just about China. In high-income countries, 40% of adults are now primarily concerned about AI and only about one in seven is primarily excited. Governments that cannot pass oversight legislation and cannot reassure their own publics have one lever left — the companies' own frameworks — and that lever is why a lab-written disclosure rule is a press release with a deadline.

What to watch

The September 24 Trump–Xi meeting is the first venue where the US could re-enter the rule-writing conversation it just left. Watch whether the midterm campaign answers the jobs number with anything other than denial, and whether Pew's next wave shows the young-adult concern curve flattening or continuing to climb — because that cohort's answer will decide which government's AI story gets believed.

If middle-income countries trust Beijing's rulebook more than Washington's, does that change what gets built — or only who gets to sell it? Tell us in the comments.

Sources: Pew Research Center — Globally, more people expect AI to cause job loss than growth · Pew Research Center — Do people trust China, the U.S. or the EU to regulate AI? · Pew Research Center — Concerns are especially widespread in high-income countries · Pew Research Center — Young adults in the U.S. are increasingly wary of AI · The Verge — AI is feared globally as the destroyer of jobs