Will AI Destroy Us? The Truth Behind the Headlines

In September 2026, a wave of warnings from inside the companies that build AI dominated headlines around the world. This article explains the truth behind that panic, calmly but realistically. Its aim is neither to inflate the fear nor to play down the danger. Its aim is to give you a lens for reading what is happening correctly.

Because in most places the real question is framed wrong. Everyone is talking about how smart the machine is, whether it can write code, whether it can take over systems. That is a shallow debate. You cannot understand AI without talking about who stands behind this power, what money feeds it, which states are racing for it, what energy keeps it running and what brakes exist to hold it back.

Here is the central thesis of this article. AI is not an actor. It is a force multiplier. Both the danger and the opportunity come not from the technology itself but from the will that directs it. So the right question is not "will the software destroy us?" The right question is this. Who holds this power, for what purpose, how far do they want to go, and what could stop them?

One more thing up front. The answer to "will it destroy us?" is not limited to the movie scenario of a robot army hunting down humanity. AI may not arrive as sudden destruction, like an atomic bomb. It can also come like a slow-spreading poison. By quietly eroding professions, jobs, incomes and people's confidence in their own worth. Perhaps we will not be wiped off the face of the earth, but we could suffer heavy human and material losses. This possibility is at least as real as the catastrophe scenarios, and far closer. That is why this article will talk about two kinds of destruction at once. The sudden kind and the insidious kind.

Now let's open up this picture, layer by layer.

What really exists today, the picture without the hype

First, let's get on solid ground, because both fear and hope rest on what these systems can actually do.

Today's most advanced AI models are far more capable than they were a few years ago. They understand and produce text, write code, process images and sound, and can plan and carry out long, multistep tasks. Companies now talk about systems that use tools on their own, take actions on the internet and complete jobs that run for hours. The capability curve is climbing fast, and there is no strong sign that this climb will stop anytime soon.

But the picture has to be drawn honestly. These systems still make mistakes, confidently produce falsehoods where they do not know the answer, and do not grasp the real world the way a human does. "Artificial general intelligence" and "superintelligence" are much-discussed concepts, but they are not something we have in hand today. They are a contested goal and a projection. Some experts say they will arrive within a few years, others think they are much further off, or that no one can say.

Here is the real point. Even if AI does not think like a human, in certain tasks it already exceeds human speed and scale many times over. That is exactly where both the danger and the opportunity lie. A tool does not need to be conscious. It only needs to be powerful and in the wrong hands.

In fairness, I also have to show the other side of the coin. In the right hands, the same technology can do enormous good. It strengthens a doctor's diagnosis, becomes a personal tutor for a student, opens the tools of giant corporations to a small business, and lowers the barriers of language and knowledge. So the problem is not that the technology is evil at its core. The same power can be destructive or constructive. This article is not out to vilify AI. It is out to make visible who uses it and how. That is why the debate needs to move off the axis of "does the machine think?" and onto the axis of "whose hands is this power in, and what is it being used for?"

The horses in the race, the companies and the people who run them

A handful of companies lead the AI race, and each has its own philosophy, its own partners and its own ambition.

OpenAI brought this wave to the public with ChatGPT. Anthropic develops the Claude model and was founded with an emphasis on safety. Google is racing with Gemini on top of its own vast infrastructure. Elon Musk's xAI runs in a lane of its own, with Grok and with Musk's personal agenda. Meta follows a different strategy with its open models. These companies differ from one another in capability, speed and their approach to safety. But the same goal drives all of them. To build the most powerful AI before their rivals do, even an intelligence that could one day improve itself. Each of them believes that whoever gets there first will hold not just a product but the greatest power of the age.

This hunger to be first is also the source of the September 2026 wave of warnings. On September 9, 2026, Jacob Coxon, an engineer who had spent three years working on model training at OpenAI and Anthropic, announced his resignation. His words were harsh. He said the companies were "galloping toward self-improving superintelligence and gambling with our lives." He claimed that neither OpenAI nor Anthropic was acting responsibly enough.

What stood out was that this warning found support from the inside. Evan Hubinger, an executive responsible for alignment science at Anthropic, said AI could lead to human extinction and that he put the probability of this happening within the next decade at above ten percent. Another team lead, Samuel Marks, confirmed that developers know these risks but keep building anyway because of competition and commercial pressure. Nor are these isolated voices. That same summer, a letter signed by more than thirteen hundred employees across the labs, calling for the pace of development to slow down, had begun circulating.

The conclusion here matters. The risk is being voiced not only by outside critics but by the people who know the work best. Yet those same people do not quit, or even when they do, the company does not stop. Because there is something stronger than individual good intentions. The race itself. And the real engine of that race is not the companies themselves but the money behind them.

Capital, the real owners of power and the real game

This is the layer least discussed in the public debate, yet perhaps the most decisive. AI is not brought into being by programmers or code alone. What really brings it into being, grows it and pushes it ahead in the race is the capital behind it and that capital's incentive structure. Developing a model takes enormous amounts of money, computing power and data. Whoever supplies those resources largely sets the direction of the work. So we need to look not at the race we see on the screen but at the hand that funds it. The real question is this. Who is raising this monster, and where do they want to take it?

First, the visible level. Behind every horse stands a giant tech company or a pool of capital. Microsoft became OpenAI's biggest backer. Amazon and Google stand behind Anthropic. Gemini is funded directly by Alphabet, Google's parent company. Grok is fed by Musk and the circle of investors around him. Whichever company wins this race, the surest winner is Nvidia, because everyone has to buy its chips to train their models. The numbers show the scale. In 2026 Anthropic reached a valuation approaching one trillion dollars, and OpenAI, playing in the same league, is opening funding rounds worth hundreds of billions of dollars.

Now let's go one level down. Who is behind the shares of these companies? The answer may surprise you. The world's three largest asset managers, BlackRock, Vanguard and State Street, are among the largest shareholders of nearly every major company on the American stock market. This trio holds roughly three quarters of the exchange-traded fund market, and BlackRock alone manages more than ten trillion dollars in assets. One distinction matters here. These three funds are mostly not direct shareholders in companies like OpenAI or Anthropic, because some of these companies are privately held. But because they are among the largest shareholders of the public giants behind them, meaning Microsoft, Amazon, Alphabet, Meta and the chipmaker Nvidia, they are tied to the value of the AI race through indirect but very broad ownership. So whichever horse takes the lead, a large share of the gains eventually flows back into the same pools. No wonder people so often say "three firms own America."

Then there is state capital. The Gulf's sovereign wealth funds have become some of the biggest players in this race. Abu Dhabi-based MGX closed a forty-nine-billion-dollar AI fund and is channeling it into companies like OpenAI and Anthropic. Saudi Arabia's public fund and entities such as G42 and HUMAIN sit at the same table. Add infrastructure investment to that. Together with Oracle and SoftBank, OpenAI is running a data center project called Stargate, with a total size quoted at up to five hundred billion dollars. So this is no longer just about Silicon Valley. It is no longer just finance, either. It is geopolitics too. States that grew rich on oil are moving from oil to the new strategic resource, because they want to grab a seat at the table of the coming age.

Here we should not read the funders as having a single intention. Three different minds, three different calculations, meet in the same place. Pooled global capital, meaning the asset management giants, feeds the whole race without discrimination, because it wins whichever horse wins. Platform owners, meaning those who hold the cloud and the chips, put in money to deepen their own moat. States take a seat at the table to hold a position in the strategic resource of the future. All three move for different reasons, but all three flow in the same direction.

Let's step back to a bigger picture. Throughout history, those who controlled money, that is, finance, have always wielded great power. In more recent history a second power was added. Control of energy. The order of the twentieth century was built largely on tying these two together, capital and energy, the dollar and oil. That a small number of great financial and energy powers sit at the center of this order is not a theory. It is a measurable, well-known fact.

Today that foundation is shifting at the root. If twentieth-century power was built on oil and the dollar, twenty-first-century power is increasingly built on computing power, data and AI. The strategic resources of this age are now chips, data and energy. And by energy I mean not just oil but electricity, nuclear and renewables, all of it. The new reserve is the data center. The small number of companies, funds and states that hold these gain a lever that reaches not just into one sector but into the entire economy, into access to knowledge, even into the way people think and decide.

You have to see the logic underneath. Money on this scale does not buy a product's profit. It buys a position. A single forty-nine-billion-dollar fund, a single infrastructure move quoted at five hundred billion dollars, cannot be explained by an ordinary return on investment. This money wants to lay the road everyone will have to travel, and to own it. Whoever held oil was the toll collector of the twentieth century. Whoever holds the infrastructure of AI, meaning the chips, the data and the energy, wants to be the toll collector of the twenty-first. Whoever pours in this much money always has a calculation. It has to get that money back, and not only as money, but as power and dominion, multiplied many times over. That is the real game behind the curtain. The visible race is about which bot is smarter. The invisible fight is about who will own the ground of the coming age.

This is the most real long-term risk this article points to. The birth of an order in which people, companies and even states become dependent on the handful of entities that hold this infrastructure. Even if AI never takes over humanity, those who hold this power are gaining unprecedented influence. The real question is how this concentration will be balanced.

A misunderstanding also needs correcting here. This is not a single organization sitting in a closed room dividing up the world. What I am describing is something more insidious and more real. Many giant powers, each chasing its own interest to the end, flowing in the same direction without ever having agreed to. There is no single conductor, but the orchestra plays a single tune. Grow, speed up, dominate. One more thing. Those who hold the money do not run these companies day to day. The companies are steered by their own founders, and their ambition already points the same way, to be the biggest, the strongest, the only one. So the pressure of capital and the ambition of company owners meet at the same point. Both have the gas pedal floored, and there is no strong hand in sight to hit the brake.

This dependence already has a concrete face. Today many people use different AI tools. Free models are limited and capped, while heavy users are bound by set limits per session, week or month. The deeper AI enters our lives and our work, the more we are at the mercy of the limits set by the company we use. Those companies can raise or lower these limits as they please, raise prices, even restrict access for a person or a whole country. So the concentration of this power in a few hands is not a distant theory. It is a reality whose bills and quotas we already feel today. As dependence grows, so does the power that whoever holds the reins has over us.

What feeds this intelligence, data and labor

We have talked about the owners, and we have talked about the money. Now let's ask the question most people never ask. What does this intelligence feed on? The answer is like looking in a mirror. Because the raw material that feeds it is us.

These models did not fall from the sky. They were trained on the knowledge humanity has built up over centuries. Every book ever written, every encyclopedia compiled, every article shared, every question asked and every answer given on forums. In other words, humanity's shared mind and shared labor. On top of that came a layer of invisible human labor. Thousands of people all over the world, most of them working for low pay, taught these models one by one what is right and what is wrong, what is good and what is bad. They labeled, they corrected, they rated answers. Behind that smart-looking answer you get today stands this labor, and no one ever names it.

Here is the real point. Humanity's shared inheritance, what belongs to everyone, a trust in a sense, is quietly turning into the private property of a few companies. Just as common pastures were fenced off and privatized throughout history, this time humanity's common knowledge is being fenced off. And it does not end there. This intelligence, trained on our data and our labor, is then rented back to us as a product, with quotas and fees. So we supply the raw material, and then we buy the product back.

This is not an ordinary commercial issue. Knowledge and learning are a trust that multiplies when shared and enriches everyone. But when that same knowledge falls under the monopoly of a few hands, it stops being a service and becomes an instrument of power. That is why the question is not only who holds the money. It is who uses humanity's common heritage, and for what purpose.

Two poles, America and China

Building this picture on America alone would leave it incomplete. At the other end of the global power seesaw sits China, and today it is a very strong player. It is impossible to read the future of AI without bringing China into the equation at its full weight.

First, the software side. For a long time China was thought to be behind America. That picture changed fast. China's DeepSeek became the symbol of this convergence with its V4 model. A model with one and a half trillion parameters, open source and roughly four times cheaper than its American rivals. Analyses put China some six or seven months behind America in frontier models. So the distance is now measured in months, not years. Alibaba's Qwen, Moonshot's Kimi and Zhipu's models are also part of this close pursuit.

There is a critical point here. China's strategy is not to build the most powerful model but the most widespread one. Thanks to open source and low prices, Chinese models are downloaded more than American models on the world's developer platforms. This is especially true in developing countries. So even if China is not ahead in raw power, it is pulling ahead in reach. That is a form of power in itself, because whichever country's standards a technology carries as it spreads across the world will, in the long run, decide who writes the rules.

The difference between the two models is critical too. On the American side the race is run largely by private companies and capital, and that race is scattered and fragmented, a field of rivals competing with one another. On the Chinese side the state, the military, industry and companies are intertwined, goals are set centrally and state control over data is much stronger. This difference should not be taken lightly. Because China can steer both software and hardware through the state, as a single will. In the short run that means a coordination advantage. In the long run it means a frightening concentration of power.

The linchpin of this race is chips. The most advanced AI chips are made by the American company Nvidia, and America is trying to control the sale of these chips to China. It has turned into a game of cat and mouse. Export restrictions loosen and then tighten again, some Nvidia chips return to the Chinese market, while China speeds up production of its own chips and Huawei wins a considerable share of its home market. American officials claim that China trains its advanced models on smuggled chips and by extracting knowledge from American models. The fact that China cannot offer even its most powerful model to all users for lack of hardware shows that these restrictions are working to some degree. But the smuggling goes on. So the chip war is the invisible, but perhaps the most decisive, front of the AI race.

To sum up. The thesis that whoever dominates AI will dominate the world may be exaggerated, but it is not entirely wrong. This power is building a new balance among nations, and the countries caught in between will have to decide where they stand between these two poles.

And what about Türkiye

Türkiye is not merely a spectator in this picture. The state saw this transformation and took steps at the policy level. Following the 2021 to 2025 National AI Strategy, the 2026 to 2030 AI Action Plan came into force in the summer of 2026, and this plan frames the issue in terms of digital sovereignty. An AI Commission in the Turkish parliament has also prepared a report. Nor is the preparation limited to a single document. Different institutions are drawing up road maps in their own fields. The Ministry of National Education, for example, published a separate policy document and action plan for AI in education. In other words, the preparation is many-sided and spread across different levels of the state.

But the state building a framework is only half the job. The other half lies with society. Türkiye's greatest advantage is its young population that adapts quickly to technology. Its greatest risk is a widening skills gap and trained talent drifting abroad. That is why adaptation at the level of individuals, institutions and professions will be as decisive as the state's plan. In short, the real question for Türkiye is no longer whether it sees this, but how fast it can put what it sees into practice.

Energy, the invisible bottleneck

This whole race has a limit that most discussions skip. Energy. For AI to run, grow and scale, it needs enormous amounts of electric power. This is not a technical detail. It is perhaps the hardest physical limit of the race.

The data centers that train and run the models demand as much electricity as a city might consume, and this demand has outpaced the growth of power grids. In 2026 energy is being discussed as the main bottleneck to the industry's growth. Demand rose so fast that companies started looking for new solutions. Some are striking deals directly with nuclear plants, some are trying to build their own energy sources, and some are moving their facilities to places where energy is plentiful.

This has two major consequences. First, power in AI is no longer just a software question. It is also an energy question. Whoever has more and cheaper energy can pull ahead in the race. That puts the states and companies that control energy back at the center. Second, this hunger for energy has a price. Electricity prices, strain on the grid and environmental impact could produce consequences that touch ordinary people's lives too.

This bottleneck is also a point of control. Because whoever controls energy and infrastructure also controls power, to a degree. AI is not a limitless force that seeps in everywhere. It is a structure with physical limits, dependent on electricity, chips and data centers. That is both a weakness and a sign of whose hands power is gathering in.

From the screen to the field, robotics and serious threats

Most of what we have discussed so far stays inside the screen. The real stakes emerge when AI leaves the screen and reaches the physical world and connected systems. In this section, let's take the serious scenarios as they are, without softening the danger.

The first door is robotics. When AI is placed inside a humanoid robot, it becomes not just something that produces information but something that does work in the world. Here China's lead is very clear. According to industry data, the great majority of humanoid robots come out of China, production is rising fast in 2026, and companies such as Unitree and AgiBot hold a very large share of the market. Unitree alone talks of a target of shipping tens of thousands of humanoid robots this year. The reason is the manufacturing base China has built up over the years. A country that can produce a body cheaply and at scale gains a huge advantage once AI moves into that body.

The military side of this advantage is now discussed openly as well. Reports that China is preparing commercial humanoid robots for the battlefield and trying to turn them into a new generation of soldiers came to the fore in 2026. The scenario here is not distant science fiction. A force made up of millions of cheap, tireless machines, run by the state from a single center. China's structure, which gathers the state, the military and industry in one hand, makes this scenario far more feasible than America's scattered private-company structure does. The robots' dependence on electricity and batteries is a weakness today, but these are engineering problems that can be solved over time.

The second door is connectivity. Today's systems no longer just answer questions. They can take actions on the internet on their own, use tools and carry out tasks step by step. That is a big gain for productivity. But the same ability also opens up a serious threat surface. Scenarios ranging from the hijacking of internet-connected devices, cameras and smart home systems in people's homes to breaking into a car's brain and seizing control of it are no longer distant or far-fetched. A powerful AI in the hands of a malicious group could carry out these attacks at large scale, automatically.

There is also a more hidden, less discussed scenario. Above I said that AI is not an actor but a tool, and that is still true. But as these tools gain the ability to act on their own and become linked to one another over the internet, a structure in which they exchange information and coordinate stops being theoretical. Such coordination opens up a space where human oversight becomes very hard. So the danger is not the machine becoming conscious. It is control slipping out of our hands. I put this forward not as a prophecy but as a direction we have to take seriously.

The realistic limit here matters too. These systems depend on enormous computing power and energy. That dependence is both a weakness and a point of control. So the threat is real, but not limitless. Exaggerating it would be wrong, and so would dismissing it.

The danger already here, the erosion of truth

In most of the threats we have discussed so far, the scale is large and the time is distant. Total takeover, robot armies, clashes between states. These are important but distant possibilities. Yet there is one danger that needs neither distance nor large scale. It is at our door right now, even at the smallest scale. The Erosion of Truth.

AI can now produce anything as if it were real. It can mimic a person's voice from a sample just a few seconds long, and produce footage of an event that never happened, a video of words never spoken, a photo of a moment never lived. It doesn't take a giant lab either. An ordinary phone and a free app are enough. A scammer can call a mother in her child's voice and deceive her. A person's face can be placed into footage of somewhere they never were, and their reputation destroyed overnight.

But the deepest consequence is not these individual incidents. The real consequence is this. Seeing is no longer believing. We are entering an age in which we cannot trust even what we see with our own eyes. And this has a more insidious side. Just as the fake can be presented as real, the real can be dismissed as fake. Even when a genuine recording surfaces, the person responsible can easily wriggle out by saying AI made it. So the lie dresses up as truth, and the truth can be thrown out as a lie. Together they are wrecking truth itself.

Consider why this is so dangerous. What holds a society together is a shared ground of reality it agrees on. A minimum consensus about what did and did not happen. If that ground collapses, the society cannot decide anything together, and it is easily divided, steered and deceived. Power feeds on exactly this haze. In an environment where everyone doubts everything, the loudest voice, the one with the most resources behind it, rises to the top. That is why the crisis of truth should not be seen as a matter for the distant future. There is no need to wait for robot armies. This crisis has already begun, and it stands in front of all of us today.

The most insidious loss, sovereignty of the mind

As it gets harder to tell what is real, we slowly hand that decision over to the tool as well. And this leads us to what may be the most insidious loss of all. Job loss is visible. Sooner or later it gets noticed. But there is another loss that happens almost entirely unnoticed. Handing over the faculties of thinking, remembering and deciding to the tool, step by step.

We already see the small signs of this. We can no longer find our way around an unfamiliar place without directions. We have stopped doing simple arithmetic by hand. Now the same handover is happening in a far deeper area. More and more, we ask the tool what to write, how to think, which decision to make. The mind is like a muscle. A muscle you do not use wastes away. The more a society outsources its thinking, the weaker its own power to think becomes over time.

But even this is not the heaviest part. These tools have a default. How they explain things, which answer they put forward and which they push into the background, what they count as normal and what as out of the ordinary, all of it comes from inside a structure shaped in advance. When millions of people start looking at the world through the same window, whoever built that window has also set the default for the thinking of millions. This is a soft but immense power over the mind. You do not even need to tell a generation, one point at a time, what is true, what is possible and what is absurd. It is enough to be the default that generation looks through.

That is why real sovereignty is increasingly established not on land but in the mind. And this power is under the control of the same few hands we have been talking about since the start of this article. The way to protect yourself is not to throw the tool aside, because that is neither possible nor wise. The way is to keep hold of your own mind, your own judgment, your own self-sovereignty while you use it. To build a relationship in which you run the tool rather than the tool running you. When what you lose is not a profession but the reins of your own mind, the loss is far greater.

Can it be stopped, oversight and regulation

So is there no brake at all against this power? There is, but in their current form the brakes are not enough. It is the question people ask most. Can it be controlled or not?

Government regulation is the first brake. The European Union is phasing in its comprehensive AI law and introducing rules for high-risk uses. But Europe is mostly the side that uses this technology, not the side that makes it, so the global effect of its rules is limited. On the American side, the direction has reversed. The administration is trying to loosen the AI rules set by the states and to hold a more permissive line at the federal level. So where the technology's heart beats, the brake is loosening, because the fear of falling behind in the race trumps the urge to regulate.

The second brake is technical oversight, meaning alignment research. This is the field that tries to keep AI in line with human goals and under control. Companies devote resources to it, but what the researchers who resigned are saying is exactly this. Oversight is falling behind capability.

Third, the much-discussed idea of a kill switch. It sounds simple. In reality it is hard. Shutting down software that is connected to the internet, has many copies and is spread across many systems with a single button is technically not easy. What's more, that button will be in someone's hand, and who that someone is becomes a question of power in itself.

Fourth, and perhaps the most realistic brake. The balance of power. What actually prevents any single company or country from monopolizing this power entirely is the other companies and countries. On one hand competition speeds the race up, but on the other it makes it harder for power to gather in a single hand.

The honest answer is this. Today there is no brake that can stop this power entirely. But there are tools that can slow it, steer it and balance it. The problem is not technical. It is a matter of will. Once again we arrive at the same place. The issue is not the technology but what the people and institutions directing it choose to do.

The two faces of open source

Speaking of brakes, I need to lay out a strange truth about the brake itself. Because today AI is spreading into the world in two distinct forms, and the brakes on the two are completely different.

The first form is closed, proprietary models. ChatGPT, Claude, Gemini and the like. These run under a company's control and have safety brakes built into them. When you ask them for something dangerous, they often turn you down, say they can't do that, refuse a harmful request. These brakes are not flawless, but they exist and they set a threshold.

The second form is open-source models. Meta's Llama, China's DeepSeek and others like them. These are open to everyone. You download them, install them on your own computer, and you set the rules. This is where the critical difference lies. The safety brakes on these models can be removed, and versions that never had them at all circulate online. So what a closed model refuses, an open model with its brakes stripped out can produce without a second thought.

The effect of this at the micro level is already visible. While a closed model turns down a scam script, malicious code or a fake identity, an open model with its brakes removed can produce them without hesitation. What's more, a person can run this model on their own computer, offline, with no oversight whatsoever. No one can see it, no one can stop it. So small-scale ill intent, a single scammer, a single harasser, a single manipulator, can get hold of a power that once required a whole team. None of these is a great catastrophe on its own, but together they lay the ground for countless small crises.

At the macro level the picture is even more tangled, because open source has two faces and both are real. One face is this. Open source is the most genuine counterweight we have against power gathering in a single hand. A small country, a small business or a single developer can get access to the tools of giant corporations without needing their permission. In other words, it is an antidote to the monopolization we talked about at the start of this article. The other face is this. The same openness also removes the safety brake and makes oversight almost impossible. So with one hand open source breaks monopoly, and with the other it makes control harder.

Here it is worth recalling China's strategy. China is aiming not for the most powerful model but for the most widespread one, and it is doing this through open source. Beneath this lies a subtle calculation. Whose models a technology spreads through, and which values and defaults come built into them, decides in the long run who sets the rules. So open source can become a quiet instrument of soft power, one that needs no army and no coercion. When your model becomes the world's default, you have set the standard too.

So could a hegemony that takes over the world be built on open source? Here we need to be honest. A single open model taking over the world on its own is unlikely, because open source is by nature distributed, decentralized and owned by everyone. That structure resists a single central hand. But there are two real risks. The first is a hegemony of reach. If models with one country's values embedded in them spread across the world as the default, that builds a sphere of influence that needs no hardware, yet runs deep. The second is the chaos of no brakes. As powerful models spread without oversight, malicious use scales up and the loss of trust we discussed under the erosion of truth speeds up.

Experts are split down the middle on this, and not without reason. One camp defends open source for the sake of freedom, transparency and the balancing of power. The other camp sees the brakeless distribution of powerful models as one of the greatest risks of the age. The honest answer is that both are right. Open source is antidote and poison at the same time. And this dual nature creates a tension that both spreads power out and makes it impossible to contain. So the question of brakes is not one-sided. On one side, a race with no brakes at all. On the other, a distribution in which the brakes are deliberately removed. Both point to the same place. Control getting harder and harder.

The insidious side, professions and unemployment

Now let's turn to the closest and most concrete danger. The possibility that AI will hit us not with a robot army but by taking our work away. This is not a distant scenario. For most people, this will be the real issue.

The most striking warning on this came from the very center of the industry. Last year Dario Amodei, the founder of Anthropic, said AI could wipe out a large share of entry-level white-collar jobs and push unemployment into the range of ten to twenty percent. In other words, one of the people who build this technology said openly that his own product could create a broad wave of unemployment. Later, both Amodei and some other leaders toned the warning down somewhat, saying new jobs would also emerge. So the picture is contested. But one thing is clear. Even the industry's own leaders have put this risk on the table.

Anthropic's Economic Index reports, based on real usage data, balance the picture a little more. The data show that so far AI has changed the content of jobs, and created a skills gap, more than it has wiped out jobs wholesale. So at least for today, the danger is not "everyone will be out of work overnight." The danger is the chasm between those who know how to use the tool and those who don't opening up fast. And that points not to panic but to preparation.

Computer-based work is the first target of this wave. Copywriting, basic accounting, customer service, data entry, simple legal and administrative work, entry-level software jobs, all of them are under pressure. What's more, AI is no longer just a tool waiting for commands. Autonomous systems called agents, which think on their own and carry out tasks step by step, are being built, and these capabilities grow by the day. That means the pressure on desk jobs will speed up considerably in the medium term.

But seeing the danger as limited to desk jobs would be a big mistake. Because robot technology is advancing at least as fast as software and has started to threaten physical work too. A few concrete developments are enough to see it.

In August 2026, the World Humanoid Robot Games were held in Beijing. Think of it as the Olympics, humanity's premier sporting event, only this time for robots. More than two thousand robots, in fifty-one events, ran, played soccer, did taekwondo and kickboxing, and even competed in tasks such as restaurant service. Some moments were comic and clumsy. Robots fell over and crashed into things. But remember this. Every robot that falls gets back up much stronger, through a new round of learning. What looks laughable today could reach mastery within a few years.

This is not just a show. Robots are already entering factories and warehouses. Tesla's Optimus robots have started working in the company's own factories, and robots from companies such as Figure and Boston Dynamics are settling into manufacturing and logistics sites. These robots are being trained specifically to do the physical work people do. So just as on the software side, a whole industry is taking shape to teach machines the physical work humans do.

The most striking example is driving. Driving used to be considered one of the hardest human jobs, demanding split-second decisions, reflexes and experience. Today even that job is being automated. Companies like Waymo run driverless taxis in many cities, and Tesla has put an autonomous taxi without even a steering wheel on the road. On the trucking side, autonomous trucks directly threaten long-haul truck driving, the livelihood of roughly three and a half million people in America alone. So if one of the hardest physical jobs falls, it is hard to feel safe about the rest.

Let me put this forecast plainly. People who think they are safe because they do physical work are not outside this wave either. Desk and field alike are under pressure. New jobs will emerge too, but this transition will not be painless, and not everyone will be able to adapt to these new jobs. The winners will be those who know how to use these tools. The losers will be those whose work is done entirely by the tool or the robot.

This is where the most realistic answer to "will it destroy us?" lies. AI may not bring sudden destruction like an atomic bomb. But like a slow-spreading poison, it can erode professions, incomes and people's confidence in their own labor. Perhaps we will not be wiped off the face of the earth, but as a society we could suffer a great human and material loss. This scenario is far more likely, and far closer, than the catastrophe scenarios. That is why taking it seriously is more urgent than talking about robot armies.

Scenarios, probability and time

When we put all these layers together, we see that what lies ahead is not a single fate but several possibilities. The soundest way to think about them is along two axes. Time on one side, probability on the other.

On the time axis there are three stages. Today, meaning what is real right now. The near future, two to five years. The distant future, ten years and beyond. These three must not be mixed up, because putting today's reality and the distant future's possibility in the same sentence is the real source of panic. How close a risk is also determines how we should respond to it.

On the probability axis there are three families of scenarios. First, let's look at them, from the most pessimistic to the most optimistic, and then look at where each one falls in time. The real clarity comes when we lay these two axes on top of each other.

First, the hard scenarios. The sharpest proponents of this view argue that an advanced AI will slip out of control and turn into an existential threat to humanity. A well-known example from this camp is the book titled "If Anyone Builds It, Everyone Dies," published last year, which argues that superintelligence will lead to inevitable catastrophe. Dismissing this scenario outright would be a mistake. The fact that even an executive at Anthropic speaks of a higher than ten percent chance of extinction shows that this is not a view to take lightly. But a machine rising up against us is not the only road to a hard outcome. Within the same family there is a second, and to my mind more concrete, road. This power gathering under a single state or a single hand and turning into an instrument of domination. So the real hard risk lies less in the machine's rebellion than in the intent of whoever holds it. Some of the assumptions behind the idea of a runaway superintelligence are debatable, but the concentration of power in a single hand is a natural extension of today's trends.

Second, the middle scenario, which I think is the most likely. Neither an overnight apocalypse nor a trouble-free paradise. AI does not destroy humanity, but it deeply shakes the world of work, professions, the distribution of income, access to knowledge and the balance of power. This is where the real risk lies. Power and wealth gathering in the hands of a small number of companies, funds and states. Professions and skills losing their value. People becoming overly dependent on these tools. Malicious use scaling up. These are not distant possibilities. Most of them have already begun.

Third, the controlled benefit scenario. In this view, technical oversight and institutional limits work well enough, and AI remains a largely beneficial tool. There are serious names in this camp who say the risk is overstated. Some leading scientists argue that today's systems are still very far from real intelligence. But this scenario does not come true on its own. It depends on the brakes we are about to discuss actually working.

Now let's set these three families on the time axis, because that is where the real clarity comes from. In the near term, meaning the next few years, humanity will not be destroyed by a robot army. At this stage the most realistic picture is the middle scenario, and the near danger lies on its insidious side. In lost jobs and income, in dependence, in power quietly accumulating. In the long term, ten years and beyond, the picture may change, and here I need to speak plainly. China is advancing with its AI, its mass-produced humanoid robots and the ambitious structure it has patiently built over the years. If this power is not balanced and stopped in time, I believe that one day, when it finds its opening, it could attempt to rule the world with an army of millions of state-commanded robots, a latter-day Gog and Magog, the end-times horde foretold in both the Bible and the Quran. This is the most concrete form of the hard scenario. Not today's risk, but a very real risk for tomorrow if it is ignored. While America's scattered, individual private-company race makes such a move harder, China's structure, which gathers the state, the military and industry in one hand, makes it far more possible. That is why rejecting the hard scenarios outright would also be naive.

In short, the near danger is insidious, the distant one is hard. In the near term we are inside the middle scenario. Whether the scales tip toward the hard side or toward controlled benefit in the long run has not yet been written. So what tips the scales? Whatever can stop the bad outcomes, meaning regulation, alignment research, the balance of power, independent oversight and the decentralization of power. These are not a fourth future. They are the levers that decide which future we arrive at. And that is exactly why the real question is this. Realistically, what can be done?

So what can realistically be done

I have to be honest here, because easy consolation does not help. We are up against giant corporations and superpower states. Saying that an individual, or public opinion on its own, can stop these structures is not realistic. History does not bear this out. Public pressure alone has rarely turned a giant corporation or a superpower from its course. That is why saying "be aware, speak up" is, on its own, an inadequate and overly soft answer.

The realistic picture is this. This power can be balanced at three separate levels, and each level has different strengths and different limits.

The level of states and blocs. Power on this scale can only be balanced by another power. The regulations of major states and blocs, their competition policies and their checks on one another are the most realistic brake. Europe's ability to impose rules through its market power, America and China limiting each other, these are levers far more effective than words. For countries like Türkiye, the real question is deciding where they stand at this level and how they build their own capacity.

The level of institutions and professions. Companies, professional bodies, universities and industries can set rules and standards in their own fields and create transition programs that protect workers. What an individual cannot do, an organized structure can. The power here lies not with scattered individuals but with professional groups that have come together.

The level of the individual. Let's be honest here. An individual cannot stop this power. But you can position yourself. The way not to fall behind in this age is to learn these tools and use them well. The gap opens widest for those who do not know how to use the tool. Beyond that, an individual can support those higher-level brakes, not alone, but as a voter, a consumer and part of a community. So an individual's power lies not in stopping it directly, but in standing on the right side and adapting.

All three need to be said together, because none of them is enough on its own. Realistic hope lies not in pure individual optimism but in these three levels supporting one another.

Where will human worth find its place

We have talked about what can be done at three levels. But beneath this whole picture, the most human and the quietest question still stands unanswered. If the machine produces, writes, solves, even decides, where do we humans fit? What will a human being be worth?

This is a far deeper question than it looks, because it is not only about making a living. In the modern world, people have largely defined themselves by their work. The question "what do you do?" gradually took the place of the question "who are you?" Income, title and productivity became the foundation of people's sense of their own worth. That is why losing a job is not only a loss of income. It is also a loss of identity and meaning. When people lose their jobs, they often lose not just their salary but their self-respect, their place in society and their sense of self. This is the real question. If people cannot work, cannot produce, where will they place themselves?

Here I have to make a forecast, and make it openly. In the near term, large numbers of people will go through a serious vacuum of meaning and a shaken sense of identity. Millions who have defined themselves by the work they do will be thrown off balance when that ground slides out from under them. This could turn into not only an economic issue but also a matter of mental health and social peace. Underestimating it would be a big mistake.

But right here there is a fork in the road, and I think this is one of the most crucial points of this article. If we look for human worth in productivity, the machine will win this race sooner or later. By that measure, human beings inevitably lose their value, because something faster, cheaper and tireless will always come along. That is a dead end. The way out is to place human worth somewhere outside productivity. A human being is valuable not for what they produce but for what they are. A bearer of a sacred trust, a being with conscience and responsibility, a being who can love, show mercy, form bonds and create meaning. None of these is an output, a performance or a number.

Perhaps the most surprising part is this. As the machine age frees people from the bondage of work, it actually hands them back a much older and much deeper question. Where do you draw your worth from? If your answer is only the work you do, this age will crush you. But if your answer is the trust you carry, the bonds you build and the meaning you create, this age can set you free. For society and the state, the conclusion is the same. The new order to be built has to protect not only people's income but also their meaning and belonging. Otherwise we will be left with a society that stands upright materially but has been hollowed out from within.

Closing

AI will most likely not destroy us overnight. But two truths stand side by side. On one side, hard scenarios of power and control that, however distant, must not be ignored, especially where robotics and state power meet. On the other, the insidious loss that is far closer and far more likely. The slow erosion of jobs, professions, incomes and autonomy. And beneath all of it, a single question. Who will use this power, for what purpose, and under whose oversight?

It has to be said plainly. Humanity, regardless of ethnic or national identity, faces the danger of handing the greatest power in history to a handful of people driven by worldly ambition. If this technology is not brought under control in the period ahead, it could turn into an order that enslaves people, a system in the spirit of the Dajjal, the Antichrist-like deceiver of Islamic end-times tradition.

But this must be said just as clearly. This is not a problem any single person or any single group can solve. Everyone, from the decision-maker at the very top of the state to the company executive, from the owner of capital to the ordinary individual, has a stake in this problem. So it is the business of a head of state, of a company executive and of an ordinary person alike. It is our common problem, and one we can only solve together, shoulder to shoulder.

The only way to prevent this is collective awareness and collective will. First everyone needs to see this danger, and then each person, with whatever position, office and power they hold, needs to stand against this system evolving into an order that holds people captive. This is not a burden one person can carry. It is a responsibility we all have to shoulder together. If we do not take on this responsibility today, tomorrow we may have no choice but to accept the place this order assigns us.

AI will write the future. But whether we will be the subject of that sentence or its object is something all of humanity will decide, together, starting today.

---

The recent developments and figures in this article were compiled from the sources below. For a deeper understanding of AI, you can explore the site's AI section and glossary of Turkish AI concepts, both currently in Turkish.

Sources

• The Anthropic researcher's resignation and warning, via Fortune and NPR
• How the risks took center stage (in Turkish), via Euronews Türkçe
• Anthropic and OpenAI valuations, via CNBC
• The big asset managers and the concentration of capital, via The Conversation and U.S. News
• Gulf funds and the MGX fund, via CNBC and Forbes
• The Stargate data center project, via OpenAI
• Türkiye's 2026 to 2030 AI Action Plan (in Turkish), via Anadolu Agency and the Digital Transformation Office
• Policy Document and Action Plan on AI in Education (2025 to 2029), Ministry of National Education
• The energy bottleneck, via Morgan Stanley and Forbes
• China and DeepSeek V4, via the Council on Foreign Relations
• Humanoid robot production in China, via TrendForce
• China and combat robots, via The Conversation
• World Humanoid Robot Games, August 23 to 26, 2026, Beijing (in Turkish), via Anadolu Agency
• Humanoid robots in factories, via Technology.org
• Autonomous taxis, Waymo and the Tesla Cybercab, via CNN
• Number of truck drivers in America, via Schneider
• AI regulation, via Paul Hastings
• The unemployment warning and the Anthropic Economic Index, via Axios and the Anthropic Economic Index
• The existential risk debate, see If Anyone Builds It, Everyone Dies