Is AI going to make humans extinct?

This week, AI was apparently going to make humans extinct.

That was the headline, anyway.

The coverage followed warnings from a former Anthropic researcher, who said he believed there was a 10% chance that AI could cause human extinction within the next decade. Other researchers have raised concerns about companies racing to build increasingly powerful models without knowing whether they will always remain under human control.

That is clearly not something we should dismiss.

But once the word “extinction” appears in a headline, it becomes difficult to have a sensible conversation about anything else.

And there is a much bigger conversation we need to have.

We don’t all see AI in the same way

Broadly speaking, people seem to approach AI from quite different positions.

Some are genuinely frightened by it or even hate it. They see job losses, deepfakes, autonomous weapons and technology becoming more powerful than the people who created it.

Others are excited by what it could do and want development to move as quickly as possible.

There are also plenty of people who rarely think about AI at all, even though they are probably already using it in more places than they realise.

I sit somewhere between the excitement and the concern.

I am fascinated by AI. I use it most days and can see enormous potential in what it can do. But I also believe we need much stronger regulation around how the most powerful systems are being developed.

I don’t think curiosity and concern are opposites.

The more I use AI and learn about it, the more convinced I become that we need both.

AI is already part of everyday life

For many people, AI still means opening ChatGPT and typing a question into a box.

But that is only the most visible version.

AI is already built into recruitment platforms, customer service systems, fraud detection, search engines, navigation, email filters, medical imaging and the software businesses use every day.

It decides which emails are probably spam. It recommends what we might watch, buy or listen to. It helps banks identify unusual transactions and predicts the next word in a message.

We have also been helping to train these systems for years.

Every time we click a recommendation, correct an autocorrect suggestion, report an email as spam or choose one search result over another, we create information about what worked and what didn’t.

Most of us didn’t call that AI. We just thought the software was getting better.

Now generative AI speaks to us. It writes, creates images, summarises documents and responds in a way that can feel remarkably human.

That makes it more useful, but also easier to misunderstand.

Why AI fascinates me

I use AI in my business because it helps me work differently.

It helps me get through information more quickly, compare documents, test an idea and look at a problem from another angle.

It doesn’t remove the need for my judgement. Quite often, the value comes from disagreeing with what it gives me or noticing the question it hasn’t answered.

But it gives me somewhere to start.

It can turn a blank page into something I can challenge. It can spot a difference between two documents that I might miss after reading them for the fifth time. It can take a collection of thoughts and help me find the structure within them.

That is genuinely useful.

And these are relatively small examples.

In healthcare, AI-assisted chest X-ray tools are already being used as a second pair of eyes for radiologists. According to the Department of Health and Social Care, more than four million patients have received a faster lung cancer diagnosis or all-clear through these tools.

AI is also being tested in clinical decision-making, drug development, patient monitoring and matching people with suitable clinical trials.

Not every trial produces a dramatic result. That is important too. We need to know when AI improves something and when adding it to a process makes very little difference.

But what about jobs?

This is one of the biggest concerns people have about AI, and understandably so.

Some jobs will disappear. Some organisations will use AI to reduce headcount, and certain tasks that currently provide entry-level work may no longer require a person.

It would be dishonest to pretend otherwise.

But exposure to AI does not automatically mean that an entire job will vanish. A job is usually made up of many different tasks, and AI may be able to take over some of them without replacing the person doing the job.

Skills England’s 2026 report says AI is likely to automate or support parts of many occupations, changing the tasks people perform rather than simply removing every affected role. It also expects AI to create new opportunities and occupations. However, there are early signs that junior and entry-level positions may be particularly exposed.

That creates difficult questions.

If AI removes the more routine work, how will people gain the experience they need to take on more complex responsibilities? Who benefits from the time saved? Will it improve someone’s working day, or will they simply be expected to produce more?

It is too early to know the full impact on employment. That does not mean employers should wait until the picture becomes clearer.

They should be thinking about this now.

Not just which tasks can be automated, but how roles will change, what skills people will need and how employees will be supported through that change.

MorninThe people raising concerns are not the enemy

I don’t think the answer to sensationalised coverage is to dismiss the people warning us about the risks.

If somebody working inside one of the leading AI companies believes development is moving too quickly, I want to understand why.

The concern is not simply that today’s chatbot will suddenly decide to destroy humanity. It is about what happens as systems become more capable, more autonomous and better able to act without somebody approving every step.

It is also about the companies developing them.

They are competing for investment, talent, market share and the prestige of getting there first. Slowing down could mean losing ground to a competitor.

Can we reasonably expect the companies involved in the race to decide where the finishing line should be?

I don’t think we can.

That is why regulation matters.

We need clear expectations about testing, transparency, accountability and where human control must remain.

If a model is powerful enough for its own developers to discuss extinction risk, “trust us” is not a sufficient regulatory framework.

I’m still optimistic

Despite all of this, I am still optimistic about AI.

I can see what it enables me to do in my own business. I can see how it could remove hours of repetitive work from an HR team. I can see its potential in healthcare, research, accessibility, education and processes that are currently slow or difficult to navigate.

I don’t believe AI is evil.

I also don’t believe every new capability should be pursued simply because somebody has discovered that it is possible; just because we can, doesn’t mean we should.

For me, the sensible position sits between fear and blind enthusiasm.

Use it. Learn about it. Question what it gives you. Look for the work it can genuinely improve.

But also think about the people whose work will change, the decisions we are prepared to hand over and the limits we need to set.

The extinction headlines will eventually move on.

The more immediate question will remain.

Who gets to decide how AI is developed, where it is used and what happens to the people affected by it?