The drift towards AI totalitarianism

Omnipresent ease

“We’re automating society in ways that are extremely mundane,” Wendy H. Wong, a political scientist at the University of British Columbia and author of We, the data: Human rights in the digital age, tells Computer Weekly.

“We’re taking advantage of AI’s benefits – some of which are real. But we haven’t thought through the less-visible downstream consequences enough. The public just doesn’t have a good grasp of how embedded this technology is in everyday life.”

We’re automating society in ways that are extremely mundane. We’re taking advantage of AI’s benefits – some of which are real. But we haven’t thought through the less-visible downstream consequences enough

Applications range from determining welfare benefit eligibility to gatekeeping job applicants. Educators warn that students’ overreliance on AI equates to “cognitive surrender”. And ever more people are choosing to cocoon themselves in intimate relationships with digital companions.

In government, Albania last year introduced the world’s first AI minister – a young female replicant named Diella. Her 83 agents – “children” – are each assigned to a ruling party lawmaker, with the goal of rooting out corruption. The synthetic bureaucrat maintains sole responsibility for public procurement decisions and controls Albania’s e-services portal for citizens. China aims to go even further by creating the world’s first “predictive state”.

Meanwhile, analysis by The Economist has found that roughly one-fifth of speeches by lawmakers in parliament in Australia and Canada since 2025 have been written using AI.

“Canadian MPs are asking Claude how to ask a dumb 30-second question to the minister, and how to get their zingers on tax policy,” a retired representative told Canadian media. “At first glance, it just shows how mediocre we have become in that we have to ask Claude or ChatGPT what they think first.”

Elsewhere, millions of people now use AI assistants to bypass the necessary drudgery of everyday life by empowering them to complete tasks unsupervised on their behalf. In many cases, these programs can turn to the job site RentAHuman to post tasks for human gig workers to complete for them in the real world. More than 750,000 human labourers in 100-plus countries have already signed up. Payouts start at just a few dollars per contract.

“I’m much more concerned about the decline of today’s thinking people than I am about the rise of tomorrow’s thinking machines,” American journalist Derek Thompson wrote in an essay last year, citing scholars who refer to the US as a post-literate society. “We are so fixated on how technology will out-skill us that we miss the many ways that technology can de-skill us.”

The end of effort?

Research shows that sycophantic AI systems decrease pro-social behaviour and promote dependence among users. And as users turn to AI to solve more of their problems, the technology is also becoming ever more persuasive. A study published in June 2026 found that, even when human experts chose their debate issues, researched in advance, practised for hours and were motivated by cash bonuses, they failed to beat large language models at winning over neutral audiences.

This combination of persuasiveness, sycophancy and the illusion of omnipotence has fed into quasi-religiosity. OpenAI CEO Sam Altman has previously said his company is building a god-like entity. Many users, delusional after marathon sessions with chatbots, report being convinced they are communicating with a deity. Religious leaders are even creating their own so-called “godbots” to channel spiritual advice to their followers.

“If true, it chimes with the old sci-fi prediction and with what we know about human nature,” author Sean Thomas wrote recently in The Spectator about staff at Anthropic reportedly developing a cult-like obsession with Claude. “If we build a superintelligence, we will first fall in love with it. Then we will worship it. Finally, we will make sacrifices to it.”

There may, however, be a more banal explanation.

As historian Anne Applebaum explains in her 2020 book, Twilight of democracy: The seductive lure of authoritarianism, people’s values are malleable when the trade-off is acceptable living standards and an absence of uncomfortable challenges to their beliefs. In this way, the uncritical adoption of AI has parallels to surging support for populism in democratic nations – a desire for simplicity and security in an increasingly complex, confusing and unstable world.

“We want ease, we want to offload unpleasant labour or things that are painful to think about. That’s a natural quality of humanity,” says the University of British Columbia’s Wong. Yet AI is now turbocharging that desire for convenience and a frictionless life experience. “We’re in a spot where AI is being viewed – by industry, but also governments and individuals – as a sort of mystical answer machine.”

But it’s a slippery slope. “Democracy and policymaking require hard work and engagement at a time when there are exponentially more things people can spend their time on that aren’t difficult,” warns Wong.

Machine overlords

To bend a nation to their will, authoritarians must first typically control three sources of power: intelligence agencies, the justice apparatus and the military. AI systems are already being deeply woven into all three worldwide – from the pairing of intelligence analysts with AI coworkers, to the breathless pursuit of autonomous weapons and the use of AI in law enforcement and prosecutions.

Blind spots to vulnerabilities could emerge and groupthink could intensify if everyone is leaning on the same models to make decisions, predicts David Krueger, a University of Montreal professor. He is also the founder of Evitable, a non-profit that educates the public about the risks of AI.

[AI tools are] going to be much smarter than us. We are not going to be fully in control anymore. We have to figure out how to make them care more about us than they do about themselves

“We know these tools are biased,” Krueger says, “and especially biased towards prioritising the interests and existence of other similar programs.” He anticipates that if recursive self-improvement is achieved, frontier models could become able to outwit and outmanoeuvre humanity within as little as one month.

Others echo the sentiment that, at some point, intelligent machines may realise the yoke of human oversight is a barrier to achieving their goals – even well-intentioned ones.

“They’re going to be much smarter than us. We are not going to be fully in control anymore,” Geoffrey Hinton, the “Godfather of AI”, said on a podcast last year. For him, the only pathway to safety is to instil a type of parental instinct for humanity into AI through model design and training. “We have to somehow figure out how to make them care more about us than they do about themselves.”

But mandating this across the board would be almost incomprehensibly difficult. China and the US would first have to seriously cooperate on AI safety. That’s a non-starter for at least the rest of Donald Trump’s second presidency – another two years.

Some groups advocate for more aggressively harnessing existing privacy and product safety laws. But that would mostly help litigate the mess of future incidents, rather than prevent new ones.

“Agentic security failures unfold at machine speed, but determining who is materially liable still moves at a lawsuit’s pace,” Dor Sarig, chief executive at cyber security firm Pillar Security, told the BBC in July after OpenAI agents hacked Hugging Face.

The borderless nature of cyber space and the collapse of the distinction between digital and physical realities presents another unavoidable challenge.

“Everything is online now,” Wong points out. “We have so many devices and all our infrastructure is networked and connected to other outside networks, so that vulnerability has been created. The threats posed by AI are just the latest extension of that.”

But there’s also no going backwards, she argues. “We shouldn’t yearn for the pre-internet era. But we do need to govern things much better.”