← Back to the blog

How Much Disruption Can People Take?

13 min read
AIDisruptionSocietyRegulationBig TechEssay

The car had decades. The internet is still uncharted territory for all of us. And AI is already knocking. On societies that can no longer keep up, corporations that are in on it from day one this time, and why regulation is not a killjoy but a lifeline.

I work with AI every day, and let me say it up front: I think this stuff is fantastic. Which is exactly why I'm writing this. Because while I ponder which workflow to automate next, another question keeps pushing in, one I enjoy far less: how much of this can a society actually take before it loses the thread?

Uncharted Territory Since 1969

In June 2013, Angela Merkel stood next to Barack Obama in front of the press and said the sentence the internet never let her live down: „Das Internet ist für uns alle Neuland“ (“The internet is uncharted territory for all of us”). At that point, the net was older than most of the people laughing about it. The ARPANET went online in 1969, the World Wide Web in 1991. Uncharted territory, after a good 40 years.

The ridicule was enormous, and I laughed along at the time. Today I'm no longer so sure who actually missed the joke. Because Merkel didn't mean the technology. She meant the law, the politics, the institutions that still hadn't grasped what had been going on for decades. And on that, she was simply right. She just put it clumsily.

I believe we live in times when societies are simply being overwhelmed. Not by any single technology, but by the pace. By ruthless, corporatocratic structures in which state and business have grown so entangled that nobody can hit the brakes anymore, even if they wanted to. The question that has stuck with me ever since: how much disruption can people take? And how much can a society?

To put that in perspective, it helps to look back. Humanity has digested plenty of upheavals like this, from the plow to the printing press to the steam engine. But the two most recent waves are the ones we lived through ourselves, and they show the pattern most clearly. They just moved more slowly.

The Car: A Century to Digest

When Carl Benz filed his patent in 1886, the car was a loud, stinking toy for people with too much money. Twenty years later, Henry Ford built the Model T, and suddenly people who hadn't even owned a carriage before could afford a car. If you go by the innovation researcher Clayton Christensen, that was the actual disruption: not the invention, but the moment it became cheap enough to reach everyone. For the carriage makers, the distinction was somewhat academic. Of dozens of manufacturers, Studebaker was just about the only one that managed the switch to the automobile. The rest disappeared.

And then? Then society had time. Decades. Time to write traffic codes, invent driver's licenses, put up traffic lights, rebuild cities, and learn that you shouldn't drive drunk. Every one of those lessons cost lives, but at least we had the time to learn them.

And even today, more than 100 years on, there are militant car haters. People for whom it's simply too much stench and noise. You can call that hostility to progress. Or the late echo of an imposition that humans never fully digested: a ton of sheet metal rolling through a residential street at 50 km/h was never natural. We just got used to it. And that took us a century.

The Internet: Done Already? We're Just Getting Started

With the internet, things went differently than with the car, and yet followed the same pattern. It started at the bottom, with researchers and tinkerers; it was slow, ugly and unreliable. The incumbents laughed. Then it became “good enough,” and good enough beats better when it's free. MP3s sounded worse than CDs, but they were everywhere. Wikipedia was less reliable than Brockhaus, the German encyclopedia, but it was free and instantly there. Brockhaus went to print for the last time in 2014.

The crucial part, though: the internet redistributed wealth. It turned garage companies into corporations and corporations into footnotes. Publishers, record labels, travel agencies and phone companies lost; Google, Amazon and thousands of startups won. Brutal for the losers, but at least the doors stood open.

What we didn't have on our radar back then was the price society pays. Not the economy, society. Smartphones and social media have been in mass use for barely 15 years, and only now is the evidence mounting of what they do to the attention, sleep and mental health of children and teenagers. Australia has banned social media for under-16s. In Europe, we're still debating. In other words, after three decades of the internet we haven't even settled whether a twelve-year-old should take a smartphone to bed. The answer, by the way, is no. I know that. I take mine to bed anyway.

So you see: it can take a very long time before people accept a technology this alien into their lives and even halfway understand its consequences. We're still in the middle of that transformation. And the next one is already here.

AI: The First Revolution Without Revolutionaries

At first glance, AI fits the pattern perfectly. Chatbots hallucinate, make mistakes, are worse at many things than a person who knows their field. But they're cheap, instantly available and good enough. They reach people who would never have paid for a lawyer to answer that one small question, or for a tutor, or a programmer. Textbook disruption, Christensen would say.

And then the pattern breaks. At the one point that had always been the same.

In every disruption before, the incumbents were blind. The carriage makers, the publishers, the phone companies scoffed at the new technology until it was too late. Not this time. This time, the big tech companies went all in from the start. Amazon, Alphabet, Meta and Microsoft are investing between 650 and over 700 billion dollars in chips and data centers in 2026 alone, depending on how you count, more than the national budget of most countries on earth, increasingly on credit. Nvidia supplies the chips, and anyone who believes the challengers are coming from outside should take a look at the ownership structures: Microsoft holds a stake in OpenAI, Amazon bought in there as well in the summer of 2026 for 50 billion dollars, and Anthropic is funded by Amazon and Google. The “newcomers” are tied to the corporations like dinghies to a tanker.

That's partly down to the technology itself. AI is inefficient from the ground up: it devours computing power, electricity and water in quantities no garage startup can afford. Anyone who wants to move anything here has to sink billions into data centers first. The ticket to entry no longer costs a workshop or a clever idea, it costs a power plant. Which means the barrier is no longer a question of ideas but of capital, and capital on that scale belongs precisely to those who are already on top.

And there it is again, the old unnaturalness. Just as the car pushed a ton of sheet metal through residential streets, AI is now planting data centers in the landscape that draw as much power as entire countries, with their worldwide electricity consumption likely to cross the 1,000 terawatt-hour mark in 2026, roughly Japan's annual consumption. Add to that an environment that changes in months, not generations. Humans have to digest that too, and nature right along with them. Except this time we don't get a hundred years for it, we get a few.

So what follows from all this? There is no upheaval of wealth, because everything remains concentrated in a few hands. Work gets upended, ownership doesn't. When an AI agent soon takes over the work of an agency, a call center or a translation bureau, the revenue won't flow to a new middle class building something in garages. It flows, as compute time and license fees, to a handful of companies that already rank among the most valuable in the world. The first signals have long been here: the tutoring provider Chegg has lost most of its value on the stock market, questions on Stack Overflow have collapsed, translators and stock photographers feel it in their order books.

And here it pays to look closely, because history has so far been more generous than the headline suggests. The car ruined the carriage makers, yes, but it also made enormous numbers of people rich, or at least well fed: new carmakers, suppliers, road builders, gas station operators, driving instructors, whole regions that lived off traffic. The internet was no different. It cost us publishers and travel agencies, but it spread wealth widely and produced an entire generation of companies and freelancers that would never have existed without the net. Both waves produced losers, but plenty of new winners too.

With AI, it looks different so far. Here, it's mostly the already rich who get richer. The chain from chips through cloud to model is so capital-intensive that the new wealth barely trickles down; it stays at the top. And whether AI will in the end actually improve life for the broad majority, the way the car and the internet did despite all their side effects, is an open question. Maybe it will turn out that way. But I wouldn't count on it, and I certainly wouldn't build a society on it.

Why the Laws Always Arrive Too Late

You could now say: then just regulate it. That's what parliaments are for. The problem is that legislation structurally cannot win, and there are even technical terms for why.

The first is the Pacing Problem: technology develops exponentially, laws are made in legislative terms. The gap between them isn't shrinking, it grows with every wave.

The second is the Collingridge dilemma, named after the technology researcher David Collingridge, who described it as far back as 1980: while a technology is young, it would be easy to steer, but nobody knows its consequences. Once the consequences are known, the technology has grown so deeply into the economy and everyday life that hardly anything can be changed. With the car we lived through it, with the internet we're living through it now, with AI we're just getting started.

If you want a case study: in 2024, the EU passed the AI Act, the world's first comprehensive AI law. The obligations for high-risk systems were due to apply from August 2026. In July 2026, a few days beforehand, the so-called Digital Omnibus took effect and pushed exactly those obligations back to December 2027 and August 2028 respectively. The law was defanged before it ever bit. The justification: industry needs more time. The same industry that is pouring 700 billion dollars into data centers that very year.

Our laws and frameworks will never be able to keep pace. Not because the people in Brussels or Berlin are stupid or lazy. But because the system they work in was built for a world in which a technology had a century to arrive.

What Happens When Nothing Happens

When laws permanently trail behind technology, a vacuum forms. And vacuums get filled, only rarely by the right people. Let me play out the scenario I consider realistic. It sounds like science fiction, I know. But over the past thirty years, the science fiction writers have had a better hit rate than most economic forecasts.

  1. Overload. People lose their bearings, then their jobs, then their trust in institutions that visibly can't keep up anymore. Not everyone at once. But enough.
  2. Loss of control. Rules stop working because the technology is faster than any agency and the corporations are more powerful than any regulator. What emerges, in effect, is a lawless space where what applies isn't the law but the terms of service.
  3. Anarchic conditions. Disinformation, anger and social division grow while the wealth sits with a handful of companies and their shareholders. The algorithms themselves amplify this: social media optimizes for attention, and nothing generates attention as reliably as outrage. That favors the political fringes all by itself. Whether Big Tech wants this or merely accepts it is secondary, because the business model rewards reach and weak regulation, and both gnaw away at exactly the institutions that hold a society together. People with nothing left to lose behave accordingly.
  4. Hard state power. We could end up in the condition of an anarchic society that can only be subdued by hard state power: surveillance, repression, control. Quite possibly with the very AI tools that helped cause the problem. That would be the bitter punchline: the technology that overwhelmed us ends up being used to keep us quiet.

You don't have to walk this road to its end to recognize it. The first two stages are already the present.

“Let's Not Get Carried Away”

I can already hear the objections, and some of them are good. So here are the best ones, before somebody puts them in the comments.

“But there are open models.” True, and Llama, Mistral and DeepSeek are great. Except: an open model needs data centers to train and chips to run. Both belong to the same people. Open source changes who gets to read the code, not who can pay the power bill.

“OpenAI and Anthropic didn't exist ten years ago either.” Correct. But Ford didn't get his funding from the carriage makers. The new companies hang on the capital and the infrastructure of the old ones. That's not disruption, that's a franchise.

“Prices are falling, small companies benefit.” Yes, AI per query is getting cheaper. That makes it democratic as a tool. As infrastructure it remains oligopolistic, and the margin ends up where the infrastructure is.

“Concentration has always existed. Ford, Standard Oil, Google.” Also true. But until now, every wave swapped out the winners. This time they stay the same, only bigger. That's the difference between a revolution and a line of succession.

“Society is more adaptable than you think.” Maybe. With the car and the internet, it managed. It had a century for the one and thirty years for the other, and it isn't done with the second yet. Anyone who considers that adaptability unlimited should explain why it has never been tested under time pressure.

“Regulation stifles innovation, Europe falls behind.” The classic. But: innovation toward what? If the goal is for European companies to be able to buy compute from American corporations faster, then yes, regulation slows that down. If the goal is a society that survives the transition, then regulation is the innovation.

Regulate Before Force Is All That's Left

I'm no fan of bans, and I've spent enough of my professional life with compliance forms to know what bad regulation looks like. Still, on this topic I've come to a conclusion I don't entirely like myself: we urgently need comprehensive regulation if we still want to save our societies.

Not because regulation is pretty. But because the alternative is worse. Smart rules aren't the opposite of freedom, they're its precondition. Those who don't shape this democratically now will get authoritarian control later. The traffic code didn't abolish the car. It made sure we survived it without a civil war.

What that could mean in concrete terms, in brief:

  • Competition law with teeth, one that doesn't just observe the concentration in chips, cloud and models but limits it.
  • Liability and transparency for AI systems and for those who make money with them. Whoever reaps the benefit carries the risk. That's how it works for every baker.
  • Protection of children and teenagers as a fixed guardrail from the start, not as a repair job after fifteen years like with social media.
  • Europe's own infrastructure, so that our rules apply not just on paper but also where the servers are.
  • Legislation that is allowed to learn: experimentation clauses, fixed review dates, rules with expiration dates. If the technology iterates, the law has to as well.
  • A share of the proceeds for society, from taxes on what machines earn all the way to an unconditional basic income. When work becomes scarce and ownership concentrates, there's hardly a way around that kind of redistribution.

The last point I consider the most important, even if it's the most uncomfortable. I'm convinced there's no longer a way around an unconditional basic income. Not as a left-wing pipe dream, but as a sober necessity. When you see how unjust the distribution becomes year after year, it's simply the fairest thing anyone can come up with. And it's a question of stability, too. Because at its core, it's the richest taking the jobs of the poorest, just not with their own hands, but through the AI they pour billions into. The damage lands at the bottom, the profit at the top. A society that watches that long enough without pushing back becomes unstable, and unstable societies are rarely friendly places. If the state can't or won't prevent the upheaval, then it must at least provide for its people. Anything else cedes the field to those who turn fear into capital.

How much disruption can people take? I don't know. But I know we're in the process of finding out, without a safety net, while everything keeps running, with corporations at the wheel who don't even have to be evil to overwhelm us. It's enough that they're fast.

If you think that's alarmist: yes, it is. A bit of alarm would have done Angela Merkel good in 2013, too. And back then we still had 13 years.

PS: If you'd rather take this topic as a thriller: Influx by Daniel Suarez. In it, a secret bureau locks disruptive technologies away so society won't be overwhelmed. At first that sounds like the solution to this essay, which is exactly why it's the best warning of where technology control without democracy ends up.

Sources

← Back to the blog