AI! Doomsday Is Here.

Published 19 September 2026 · Sibongile Borokini (Author)

A conversation about where artificial intelligence came from, what it is already changing, what could go wrong, and what humanity should never surrender to it.

If you have used Google Maps, watched Netflix, shopped online, asked Siri a question, spoken to Alexa, received a fraud alert from your bank, or been recommended a product you somehow ended up buying… congratulations!

You have probably been interacting with artificial intelligence for years. So why does it suddenly feel as though AI has arrived at our doorstep?

Why are we suddenly worried that it will take our jobs, replace artists, make doctors obsolete, destroy education, manipulate elections, generate fake people, write everything we read and — depending on which corner of the internet you visit — eventually decide that humanity itself is the problem?

AI! Doomsday is here.

Or is it?

Because before we decide whether AI is the beginning of the end, perhaps we should ask a much simpler question:

When did this actually begin?

Wait… AI isn't new?

This is probably the first thing that surprised me.

When many of us hear “artificial intelligence”, we picture ChatGPT, humanoid robots, self-driving cars or some futuristic machine sitting in a laboratory. But the idea is much older than that.

In 1950, British mathematician and computer scientist Alan Turing published his famous paper Computing Machinery and Intelligence, asking whether machines could demonstrate behaviour that could be considered intelligent.

Then, in the summer of 1956, a group of scientists gathered at Dartmouth College in the United States for what became known as the Dartmouth Summer Research Project on Artificial Intelligence. This is where the term “artificial intelligence” was coined as the name of a new field of research.

1956! That's seventy years ago.

So no, ChatGPT did not invent artificial intelligence. AI has gone through decades of experimentation, excitement, disappointment and renewed breakthroughs.

There were periods when researchers were incredibly optimistic about what machines would soon accomplish. There were also periods when progress failed to match expectations and interest and funding declined.

Then came more computing power. More data. Better algorithms. Neural networks. Deep learning. And eventually, systems capable of generating language, images, audio, video and computer code.

The technology didn't suddenly appear. What changed was our relationship with it.

We've been living with AI without calling it AI

Think about the last time Amazon recommended something to you. You might have thought, “This website knows me.” Well… sort of.

Amazon says it has been using AI and machine learning to improve its customer experience for more than 25 years. What began with machine-learning recommendation systems has expanded into AI-powered shopping tools and robotics used in its fulfilment network.

And Amazon isn't an unusual example.

AI has quietly been working behind the scenes in areas such as:

• recommendation systems
• search
• fraud detection
• spam filtering
• speech recognition
• navigation
• advertising
• customer service
• image recognition
• logistics

For years, we interacted with AI through an interface without necessarily thinking about the technology underneath it. Then something changed.

The interface disappeared or perhaps more accurately, the interface became a conversation.

Then the machine came home

On 30 November 2022, OpenAI publicly introduced ChatGPT as a free research preview. Suddenly, you didn't need to know how machine learning worked to interact with an AI system. You could simply type:

“Explain this to me.”

“Help me write this.”

“Why isn't my code working?”

“Teach me Excel.”

“Help me understand this medical term.”

“Translate this.”

“Give me ideas for my business.”

And for the first time, advanced AI felt less like infrastructure hidden inside somebody else's product and more like something sitting beside you.

That distinction matters.

Because AI was no longer just doing things for people without them thinking about IT. People were beginning to ask the machine to do things themselves.

And that opened a door.

What happens when knowledge gets a front door?

This is where things get interesting.

Imagine someone who wants to learn programming but cannot afford a course. Someone who wants to understand a complicated document but doesn't know where to start. A small business owner who needs help analysing their sales data. A student who is struggling with algebra. Someone learning another language. Someone trying to understand how to build a website. Someone in a country where access to specialised expertise is limited.

The thing is, AI doesn't magically solve all those problems but it can lower the barrier to getting started.

The World Bank's research on AI in developing countries describes AI's potential to expand access to information and services, increase productivity and open markets. It also highlights smaller, more affordable AI applications that can operate on everyday devices and are being explored in areas such as agriculture, health and education.

But here's the part I find even more fascinating. A study of 5,179 customer-support workers examined what happened when employees were given access to a generative AI assistant.

Productivity increased by about 14% overall but among novice and lower-skilled workers, the improvement was about 34%.

The effect was much smaller among experienced and highly skilled workers. The researchers suggested that the AI system was, in part, spreading the practices of more experienced workers to newer employees.

Stop there for a second.

We usually imagine technology as something that makes the people at the top even more powerful. But what if, sometimes, AI does something slightly different? What if it distributes some expertise? What if someone with five years of experience can use AI to help someone with six months of experience get closer to that level?

That doesn't eliminate the value of experience. It changes the distance between people with different levels of experience. And suddenly, AI isn't only a story about automation. It is also a story about access.

But access to what?

This is where we need to be careful.

Giving someone access to an AI model doesn't automatically give them equal opportunity.

You still need:

• electricity
• internet access
• a device
• digital literacy
• education
• relevant data
• skills
• money
• infrastructure

The World Bank describes these foundations in terms of connectivity, compute, context and competency. AI adoption is growing in middle-income countries, but access and capacity remain uneven, with high-income countries dominating AI innovation, computing infrastructure and startup funding.

So here's another uncomfortable thought, Could AI democratise knowledge while simultaneously widening other inequalities?

Absolutely.

A person with an AI assistant and a fast internet connection is not necessarily in the same position as someone who has never had reliable internet access.

AI may lower some barriers but it doesn't automatically remove all of them.

And then AI walked into the hospital

Now let's make the stakes a little higher.

Imagine asking an AI to write you a birthday message. If it gets the message wrong, you might laugh. Now imagine asking it to help interpret a medical scan. Suddenly, “oops” isn't quite as funny. And yet AI is already being used across healthcare.

The World Health Organization says AI has significant potential in diagnosis, treatment, health research, drug development and public-health functions. At the same time, it warns that ethics and human rights need to be central to the design, deployment and use of AI in health.

The 2026 Stanford AI Index reports rapid developments in medical AI, including AI-generated clinical notes, biological modelling and AI-enabled medical devices. It also reports that the U.S. FDA authorised 258 AI-enabled medical devices in 2025.

This, creates a fascinating tension. AI can help doctors process information. It can reduce administrative work. It can assist with research. It can potentially help discover new treatments. But what happens when an AI system is wrong?

What happens when the data used to train it does not adequately represent certain populations? What happens when a doctor trusts the machine too much? And perhaps most importantly, who is responsible when the machine gets it wrong?

The doctor? The hospital? The company that built the system? The people who trained it? The regulator who approved it? Or the human who assumed that because the machine sounded confident, it must be correct?

The machine doesn't have to hate us to hurt us

This might be one of the most important things to understand about AI risk.

We don't need to imagine an evil robot. AI can cause harm without having an intention at all.

The International AI Safety Report identifies three broad categories of risks from general-purpose AI: malicious use, malfunctions and systemic risks.

Let's start with something surprisingly ordinary.

AI can be wrong. It can invent information. It can produce inaccurate code. It can give unreliable medical information.

It can confidently provide an answer that sounds perfectly reasonable while being completely incorrect. This is often called a hallucination.

And there is something psychologically dangerous about that. When a human says, “I don't know.” We know what that means. When an AI gives us a beautifully written answer in three seconds, we can forget that confidence and correctness are two completely different things.

An articulate answer is not necessarily a true answer. That distinction is going to become increasingly important.

Then there is bias

AI learns from data. And humans produce data.

Unfortunately, human history is not a perfectly fair dataset. We have created societies containing discrimination, inequality, stereotypes and exclusion. So if an AI system learns patterns from those societies, what happens if some of those patterns are reproduced?

UNESCO warns that AI can reproduce and amplify existing biases and discrimination, while also raising concerns around human rights, privacy, environmental sustainability and inclusion. This is why “the computer decided” isn't a sufficient explanation. A computer doesn't arrive from nowhere.

Humans choose:

• what data goes into the system
• what problem the system is designed to solve
• what counts as success
• what risks are acceptable
• where the system gets deployed
• who is allowed to use it

AI may automate a decision but humans designed the conditions under which that decision happens.

And then there is the darker side of AI

The same technology that can help someone write a CV can help someone create a convincing scam. The same technology that can help create educational content can generate convincing fake content. The same technology that can help someone communicate can be used for impersonation.

The International AI Safety Report documents established harms involving fraud, impersonation, fake content, privacy violations and other malicious uses. It also notes that some risks around advanced AI remain uncertain and are still being studied.

That distinction matters because not every terrifying prediction about AI is an established fact. Some risks are already happening. Some are plausible future risks. Some remain highly uncertain. Treating all three categories as though they are equally proven doesn't help anyone understand the problem. In fact, it can make the conversation worse.

Because if everything is called “doomsday”, eventually nothing is.

But what about our jobs?

Ah.

The question everyone asks.

“Is AI going to take my job?”

The honest answer is more complicated than a simple yes or no.

The International Labour Organization's 2025 analysis estimates that one in four workers globally are in occupations with some degree of exposure to generative AI. But exposure does not mean replacement.

The ILO concludes that because most jobs still contain tasks requiring human input, job transformation is more likely than complete redundancy for most occupations.

That doesn't mean nobody will lose work. It means the story is more complicated than:

«AI arrives → humans disappear.»

Some tasks may be automated. Some jobs may shrink. Some jobs may change. Some new jobs may emerge. And some workers may become dramatically more productive.

The bigger question might therefore be, can people adapt quickly enough to a labour market that is changing faster than traditional education and training systems?

Because technology can change overnight but human beings need time.

The invisible side of AI

Now, let's step away from the screen.

AI can feel weightless. You type something, an answer appears. It feels almost magical! But reality is, AI is not weightless. Behind that answer are data centres, servers, chips, networks, electricity and cooling systems.

The World Bank describes computing infrastructure — including chips, servers, data centres and cloud services — as a fundamental part of AI readiness. Stanford's 2026 AI Index documents the rapidly expanding infrastructure and investment required to develop and operate increasingly capable AI systems.

So the next time someone says, “It's just an app.” Remember, it isn't just an app. It is an enormous physical infrastructure hiding behind a very small screen.

And that means the AI conversation is also an energy, environmental and resource conversation.

So who is supposed to keep all of this under control?

Now we're getting to the real question.

Because it would be convenient to say, but I “The technology companies should handle it.” But why should the people building the technology be the only people deciding the rules for it?

Then perhaps governments should handle it. But governments don't build every AI system. Researchers understand the technology. Companies control much of the infrastructure. Workers experience its effects. Teachers see what it does to education. Doctors see what it does to healthcare. Parents see what it does to children.

Ordinary people experience it in their everyday lives.

So perhaps responsibility cannot belong to one group. It has to be shared.

Governments have a role

Governments can create laws, standards, regulatory systems and public institutions capable of dealing with AI.

This isn't some distant conversation happening only in Washington, Brussels or Silicon Valley. South Africa is already having it.

In 2026, South Africa's Cabinet approved publication of a draft national AI policy for public comment. The policy is intended to address both the benefits and risks of AI and includes pillars around skills, inclusive growth and jobs, responsible governance, ethical and inclusive AI, and human-centred deployment.

That means AI governance isn't merely a futuristic debate.

It is already becoming part of the decisions countries have to make about their economies, education systems, workers and public services.

Companies have a role too!

The organisations building and deploying AI have enormous responsibility. But responsibility needs to mean more than, “We added an AI disclaimer.”

The U.S. National Institute of Standards and Technology's AI Risk Management Framework identifies characteristics associated with trustworthy AI, including reliability, safety, security, resilience, accountability, transparency, explainability, privacy and fairness.

In other words:

Build it. Test it. Monitor it. Understand its limitations. Give people ways to challenge it and don't pretend that a system is trustworthy simply because it is impressive.

The OECD's AI Principles similarly emphasise human rights, fairness, privacy, transparency, human oversight, safety and accountability. UNESCO's global Recommendation on the Ethics of AI also places human dignity and human rights at the centre of AI governance, alongside transparency, fairness, environmental sustainability and human oversight.

Notice something?

Different organisations. Different countries. Different frameworks. But a recurring idea...

Humans must remain responsible for what humans build.

But what about us?

This is the part we don't talk about enough.

We like to imagine that responsibility belongs to billion-dollar companies and governments. But we use these systems too!

We decide whether to believe the answer. We decide whether to verify it. We decide what personal information to put into it. We decide whether to use AI to learn or simply to avoid learning. We decide whether to let it write every thought we have. We decide whether to treat it as an assistant… Or an authority.

And that distinction might determine more about our future than the technology itself.

What should we never outsource?

This is the question I keep coming back to.

Not, “How do we stop AI?” because AI isn't going away. And perhaps that isn't even what we should want.

AI can help people learn. It can help doctors. It can help researchers. It can help businesses. It can help people with disabilities. It can help people communicate. It can give someone who has never written code a place to begin. It can give someone who has never had access to a consultant an opportunity to ask questions. It can make certain forms of expertise more accessible.

The World Bank's research captures this possibility while also warning that access, infrastructure, skills and policy will determine how widely those benefits are shared.

So perhaps the real challenge isn't stopping AI from becoming powerful. Perhaps it is deciding what humans should remain responsible for even when machines become incredibly capable.

Should we outsource judgements? Should we outsource morality? Should we outsource responsibility? Should we outsource curiosity? Should we outsource creativity? Should we stop learning because something can give us the answer? Should a doctor stop questioning because an algorithm sounds certain? Should a teacher stop teaching because an AI can explain? Should a business stop employing people simply because automation is cheaper?

And perhaps the biggest question,

If AI becomes better at producing answers, will we become worse at asking questions?

Maybe this is what “doomsday” actually looks like

Maybe it isn't a robot waking up one morning and deciding that humans are its enemy. Maybe it is much quieter.

A student stops learning because AI can do the assignment. A worker stops developing skills because AI can do the task. A patient trusts a machine without understanding its limitations. A company values efficiency more than human dignity. A government adopts technology faster than it can regulate it. A person believes something simply because an AI said it confidently. A society gradually forgets that convenience and progress are not always the same thing.

That possibility is much less cinematic than a robot apocalypse, but perhaps it deserves more attention. Because technology doesn't have to become evil to change us.

We only have to become careless with it.

So… is doomsday here?

Maybe. Maybe not. We simply don't know enough to make some of the biggest predictions with certainty.

The International AI Safety Report itself acknowledges significant disagreement among experts about advanced AI capabilities, risks and how effectively those risks can be mitigated. At the same time, it distinguishes those uncertain future possibilities from harms that are already documented today.

And perhaps that's exactly where we should leave the conversation. Not terrified. Not blindly optimistic. Awake.

Because AI is neither a saviour nor a villain sitting somewhere inside a server. It is a technology created, trained, deployed and controlled by people. Which means the future of AI isn't entirely about what machines become.

It is also about what we decide to become alongside them.

The question was never simply, “Will AI replace humans?”

Perhaps the better question is, “What parts of being human are we willing to give away?”

Maybe that, is where the real responsibility begins. Because if we want AI to benefit humanity, we cannot simply ask, “What can AI do?” We have to keep asking, “What should it do?” And even more importantly, “What should it never do without us?”

Sources