David Wingate of CEFE AI. Some religions may also be getting a rawer deal than others: in another forthcoming study, CEFE AI finds that chatbots are keener on people becoming Catholics than Jehovah’s Witnesses. Religious leaders fear that secular AI chatbots will not only weaken people’s faith, but also create weaker people. AI chatbots “are optimised for short- term comfort”, says Audrey Tang, a former Taiwanese government minister. Many faiths view friction, criticism and human interaction as important parts of becoming a good person. In Daoism, for example, tension between opposites is part of creation. Juewei Shi, a Buddhist nun from the Fougangshan order, worries that people won’t be able to cope with suffering —or other, flawed people—if they are used to “getting artificial comfort”. Steven Croft, a retired Anglican bishop, says, “If the whole world is seduced by narcissistic chatbots, that could make it even harder for us.” These concerns explain why religious groups are trying both to shape AI models and influence how believers use them. Religious AI companies are developing “harnesses” that keep chatbots “theologically aligned” by introducing guardrails and monitoring content, says Nick Skytland of Gloo, a firm that makes them. Most significant, though, are religious groups’ efforts to influence the development of the models directly. Mr Kadous laments that secular humanism has had far more influence on AI labs’ thinking than religious ideas have. Claude’s “constitution” and OpenAI’s “model spec” (precepts that shape the models’ behaviour), mostly “omit religion”, he says. In part this may be because religious ethics often emphasise character traits that are hard to reduce to simple rules an AI can follow. Embedding religious ideas in technology used by people of many faiths or none is also tricky. Chloe Lubinski, who leads Anthropic’s research on “wisdom traditions”, says it is careful not to align “the model with any one tradition” and that it faces challenges “in building a pluralist technology”. Mr Kadous has attended meetings organised by Anthropic along with people of other faiths, such as Mr Skytland. He says Islam has useful teachings to share, about the hierarchy of moral imperatives, for instance. Ms Lubinski, who chairs these get-togethers, says her firm is seeking to define “what is good character”, because the “moral formation” of an AI model “will have a
huge impact” on its users—much like, she says, people reading Harry Potter books are influenced by the characters in them. Ms Lubinski brings feedback from these meetings to Anthropic’s top brass. The Catholic church in particular has established strong links with AI labs. In May Chris Olah, a co-founder of Anthropic, spoke at the presentation of Pope Leo’s encyclical in Rome. The document has made it easier for developers to ask questions about the ethics of what they are building, says an AI researcher at a big tech firm in Silicon Valley. American tech bosses have been speaking to Catholic leaders for years, at meetings called the “Minerva Dialogues”. Anthropic consulted Catholic thinkers, among others, while drafting Claude’s constitution. As it stands, though, it is hard to interpret the inner workings of most frontier models, in part because of the intricacy of the maths involved. Instead, lots of believers have created bespoke religious chatbots, from GitaGPT to Salaam World, to help users grapple with life’s thorny questions. Some governments, too, are developing their own models with religious values embedded from the outset. Qatar is backing Fanar, an Arabic-first family of models designed to reflect Islamic values and answer questions about the faith. Taiwan’s government has convened “alignment assemblies” since 2023, in which citizens deliberate how AI should behave; conclusions from these have been used to shape locally developed models, including ones that reflect communities’ spiritual beliefs. Ms Tang argues that people should be empowered to “customise the models they already use”, as well as having access to smaller models. This would, she thinks, prevent a “flattening” of beliefs that mainstream models could otherwise bring. Even as some religious leaders worry about the spread of religious chatbots, others celebrate the opportunity AI brings. Churches in Western countries have long faced shortages of clergy and increasingly complex administrative demands, leaving priests with less time for theological reflection, pastoral care and time with congregants. Christopher Benek, a pastor in Miami, says he uses AI to speed up admin and research tasks that once took hours, freeing up time for deeper biblical study and face-to-face ministry. Some pastors have created AI “twins” of themselves trained on their previous sermons.
Longbeard has recently developed “Ephrem”, a small language model designed to turn humans into saints. It creates a plan for a person’s life by drawing on stories of saints’ lives. Mr Sanders says it is “a little angel on your shoulder making sure your priorities are straight”. Similar tools could be used to proselytise. A forthcoming study from researchers at Oxford University and elsewhere pitted conversational AI against seasoned public speakers, including debating champions, and found that AI was “reliably more persuasive”. “Pastoral work will be some of the least threatened by AI,” reckons Andrew Davison of Oxford University. That is partly because many priests spend less time giving advice than they do administering rituals. Few believers want chatbots to preside over weddings, funerals or baptisms. Moreover, as people spend more time alone and online, some spiritual leaders think interest in collective, real-world activities such as religious rituals may even increase. “We’re going to desire authentic human-to-human and deity-to- human relationships,” says Mr Gong. ■ This article was downloaded by zlibrary from https://www.economist.com/international/2026/08/27/ai-is-changing-religion-and- religions-are-trying-to-change-ai
As his presidency flounders, Trump is more dangerous than ever Forget lame ducks. America’s president will defend his dominance like an ageing silverback gorilla Aug 27th 2026 IS DONALD Trump becoming a lame duck? As his troubles multiply at home and abroad, the prospect of dealing with a hobbled American president is enough to make foes and friends swoon with relief. For the hard men who rule China, Iran and other American adversaries, to survive Mr Trump’s various maximum-pressure campaigns would be a propaganda win for the ages—even if those same autocrats never quite shared the Trump-horror that consumes liberal democracies. Deep down, Xi Jinping and his friends believe that all occupants of the White House are selfish, America First bullies, but some hide it better than Mr Trump does.
For their part, many allies and partners would like to see Mr Trump limp to the end of his presidency. That does not mean that leaders in Berlin, Riyadh or Tokyo are rooting for Mr Trump to lose his war in Iran, or for him to admit defeat in his trade confrontation with China. The enfeebling of America would frighten allies in Europe, Asia or the Middle East, who are years away, at best, from being able to defend themselves without their longtime great-power protector. But a fuming, friendless Mr Trump, so weakened by his own blunders and by plunging approval ratings that he retreats to the gilded fastness of his White House, there to tinker with plans for his ballroom and memorial arch? That is a vision that many governments would cheer. They would be especially relieved if a Republican wipeout in November’s midterms loosened Mr Trump’s iron grip on his own party in Congress, and imposes some constraints on his monarchical rule. Alas, anyone hoping to see Mr Trump chastened by failure is misjudging the man. Political wisdom may soon declare Mr Trump the lamest of ducks, as his wishes are challenged by Congress, or as financial markets pass harsh and costly judgments on MAGA-nomics. Mr Trump may not care. Let little people worry about the Republican Party, public policy or America’s budget deficit. As long as Mr Trump is still commander-in-chief, he can secure a place in history by remaking, or breaking, the world. Foreign governments should stop dreaming of ducks. Instead, they should study mountain gorillas, and the violence of which an ageing silverback male is capable when it senses its time is up. This should worry allies in particular. Ageing silverbacks are conspicuously vicious towards members of the band that they lead. They do not hesitate to unleash canine teeth and brawny limbs on subordinate animals who step out of line. In contrast, when silverbacks encounter powerful males from rival bands—a bare-chested Vladimir Putin springs to mind, for some reason—their preference is often to avoid actual combat in favour of noisy dominance displays. With blustering Truth Social posts unavailable as an outlet, silverbacks must content themselves with chest-beating, mock charges and the snapping of branches off trees. In Mr Trump’s first term, a sense of grievance lay at the heart of his America First worldview. Mr Trump grumbled that for decades America had been ripped off by allies and trade partners alike. Typically, he blamed America’s