regulation. Indeed, surveys point to overwhelming Chinese optimism about AI. In a poll by Gallup, 93% of respondents in China said it would mostly help their country, versus just 36% in America. Calls for China to work with America to slow down frontier models have been brushed aside in Beijing. “Fearmongering” serves no one, says China’s foreign ministry. In fact, China spends an awful lot of time thinking about AI safety. Telephone fraud is just the tip of an iceberg of anxieties. China’s concerns run from criminal activity, labour displacement and national security to, yes, the loss-of-control scenarios causing sleepless nights in Silicon Valley. This should not be surprising. The Communist Party, which obsesses about danger in almost every domain of life, has fought for years to bring cyberspace under control. Is a regime that views the internet as an existential challenge suddenly about to let AI rip? If, as in the Trump administration, one defines regulation as intrinsically a form of slowdown, then China is, in effect, slowing AI—though it would never say so. “China already has the most stringent AI rules anywhere in the world,” says Kendra Schaefer of Trivium, a consultancy based in Beijing. The foundational document is China’s framework for the governance of AI safety. A second update came out in September. It lists 54 risks under three categories: “inherent” (models behaving in unintended ways), “application” (use by criminals) and “secondary” (social alienation). It also proposes ways to eliminate or, at least, mitigate these threats. One is to require checkpoints at which humans (remember those?) must review and, if necessary, halt AI agents before they roam too widely. Some of the framework’s earlier ideas are already in force—for instance, AI- generated content such as audio and video must be clearly labelled. Others are likely to form the basis of new rules. Draft measures call for service providers to shut down any feature with serious problems immediately. On October 16th China also proposed curbs on exports of its most advanced AI technologies. As AI has grown more powerful, China has paid more attention to doomsday possibilities. The original version of its safety framework, issued in 2024, made just one glancing reference to loss of control. In the newest

version, that topic comes up more than a dozen times. Even so, it is possible to discern a philosophical gap. Whereas some in America fear that AI models will eventually surpass human intelligence and take control of their own evolution, the more common view in China is that AI is a transformative technology that can be harnessed and directed, not something that will slip human grasp. Xi Jinping compares AI to the steam engine, electricity and the internet— technologies that were epoch-shaping but did not pose dangers that required a co-ordinated slowdown. With China still trailing America at the frontier of AI, Mr Xi is all the less inclined to agree to apply the brakes. As China sees it, such an agreement looks both unnecessary and self-serving for Americans. Unsurprisingly, Chinese officials devote much bandwidth to ideology. They want to ensure that models do not contradict the party on important matters, and that new tools do not undermine its control. Yet there are more banal concerns, too, reflecting the soft paternalism of the Chinese state. Hence Mr Guo’s three golden seconds on the phone and a ban on offering virtual lovers or, for that matter, virtual relatives to minors. China’s yen for rules may have an unintended effect. In America, where the government has a much lighter touch, companies have stepped into the breach, evaluating their own models and asking for more guardrails. In China firms are often more passive, waiting for directions from the party and government before applying controls. A review by Concordia AI, a consultancy in Beijing, of ten Chinese firms with recently released models found that just five had published safety evaluations. That suggests possible perils if models develop faster than regulators expect. China’s rules will also fail to impress those who view its open-weight models as unavoidably dangerous. Critics fret that such models, unlike the closed ones dominant in America, can be abused more readily. Yet many in Chinese AI argue the benefits outweigh that hazard: they note that China’s open models have helped defend computer systems from rogue attacks launched by American closed models, as when OpenAI tools breached Hugging Face, an AI platform. In any case, China is already working on

tougher rules to govern open models. Insiders say that new releases face exacting safety reviews before being offered to the public. By most measures China’s AI models are a few months, or more, behind America’s in their performance. That is a big gap for such a fast-evolving technology. But do not accuse China of lagging behind America in its management of risks. When it comes to rules, controls, limits and warnings, China is in its element.■ This article was downloaded by zlibrary from https://www.economist.com/china/2026/10/19/does-china-really-prize-ai-speed-over- safety

· United States

Republicans’ identity crisis in Texas The midterms’ fundraising figures tell a surprising story Why Gen-Z conservatives love Richard Nixon The curious, predictable case of the fake poll How to share the wealth from AI? Alaska offers lessons City-hall socialists try to make populism pragmatic Donald Trump should love the new documentary about Elizabeth Holmes

United States · United States | Midterms maths

Republicans’ identity crisis in Texas The old campaign messages aren’t working and the new ones are risky Oct 22nd 2026 REPUBLICANS IN AMERICA’S biggest red state don’t usually have to play defence. But less than two weeks before election day, they are panicking. Ken Paxton, the scandal-plagued Senate candidate, is trailing James Talarico in the polls. Since early September, national Republicans have poured about $200m into ads in a bid to save his campaign—and their Senate majority. Greg Abbott, Texas’s three-term governor, has won past elections by double digits, but this year is being forced to fend off a Democratic challenger, Gina Hinojosa, outspending her six-to-one over the summer. “We’re worried about Texas,” Donald Trump admitted early this month. Mr Paxton is a uniquely bad candidate, dogged by investigations and marital infidelity. But he is not Republicans’ only problem in Texas. Latinos in

South Texas, who swung to Mr Trump in 2024, now appear to be drifting left; soccer moms in the suburbs of Dallas and Houston are tired of the president’s rancour. After more than two decades of unbroken rule, Texas Republicans are struggling to find a winning message—old campaign lines aren’t working and the party is trying to whip up the base by questioning if foreigners belong. Even if the party hangs on to power in Texas on November 3rd, its grip is loosening. Like Republican candidates elsewhere, those in Texas are being dragged down by Mr Trump. The president’s statewide approval rating is -18, according to The Economist’s tracker. Mr Paxton told donors in September that Mr Trump’s midterm convention in Dallas sank his numbers, according to a leaked recording. Mr Abbott was conspicuously absent from a Trump rally in San Antonio on October 7th. Texas’s political identity has long been secure enough to withstand a national leftward tilt. Now Republicans’ Lone Star arguments are losing force. For years, calls for border security united Texas Republicans and won over many independents. The party can now boast that the problem has been fixed—border encounters in August were 94% below the average month under Joe Biden. The issue is also less salient. For the first time in a decade, voters rank the cost of living above immigration as the “biggest issue facing the state”. The ferocity of Washington’s deportation crackdown doesn’t help. “Stephen Miller and the guys calling shots in DC have taken immigration policy from being a net positive for Texas Republicans and made it neutral or negative,” says Bill Miller, a Republican lobbyist in the state. After a federal agent killed a man in Houston in July, Mr Paxton offered no condolences. When the going gets tough, Texas Republicans tend to fall back on bragging about the “Texas miracle”, the economic boom unleashed by low taxes and light regulation. In the five years to 2025, at least 184 companies moved to Austin, Dallas and Houston. When Google announced a $40bn investment last November, Mr Abbott declared the state the “headquarter of headquarters” and promised to make it the “epicentre of AI development”.

Those words now haunt him. This year Texans, like Americans elsewhere, turned against data centres, forcing Mr Abbott to ask the legislature to repeal the tax breaks that helped lure them. Ms Hinojosa, his challenger, pounced on the hypocrisy. Campaigning in towns such as Amarillo and Granbury, where Democrats rarely win, she casts herself as a populist. “Greg Abbott is selling you out,” an ad reads. The governor’s other big policy priority, universal school vouchers, is also unpopular among Republicans in small towns, where public schools are a top employer. “The classic arguments have run out of steam and a new order hasn’t been born yet,” says James Henson, an Austin pollster. To fill the void, some Republicans are testing a new message that they hope will resonate in November and beyond: “keeping Texas, Texas”. The meaning of that slogan has recently come into focus, as politicians wage war on foreign influences they say threaten Texas from within. First it was Muslims. Last year Mr Abbott tried to block a mosque in Plano from building a community he dubbed “Sharia city” and designated America’s biggest Muslim civil-rights group a terrorist organisation. Mr Paxton tried, unsuccessfully, to bar Muslim schools from school vouchers. In the Republican primary Mr Paxton beat John Cornyn, the incumbent senator, with the help of an endorsement of Mr Trump and a new line of attack, with ads accusing Mr Cornyn of having a “special place in his heart for radical Islam”. Then it was legal immigrants. In January, Mr Abbott froze H-1B visa applications at state agencies and universities. Four months later Mr Paxton opened investigations into 30 firms he accused of using “ghost offices” to allow “foreign nationals to invade Texas”. In September Bo French, a Republican running for railroad commissioner (a job that actually involves regulating oil and gas), posted a photo of students celebrating a University of Texas football victory on X—nearly all of whom had brown skin. “The problem is now obviously far worse than anyone imagined,” he wrote. Old- guard Republicans whom the party has ostracised, including Karl Rove and Mr Cornyn, called the post racist. Texas’s top brass didn’t condemn it. The fight over who belongs in Texas is most heated in the suburbs. There, dizzying growth has transformed Republican strongholds into political