WEALTH October 17, 2026 0 Comments

Gavan Reilly Exposes the Silent AI Threat Every Entrepreneur Must Master Before It’s Too Late

Ever had a chat where the question suddenly flips your world upside down? Imagine being a government minister, calmly grilled not about budgets or policies, but the very real possibility of humanity’s extinction—yep, extinction, like we’re all extras in some sci-fi flick gone wrong. That’s exactly the tightrope Enterprise Minister Peter Burke is walking these days. When AI was just a clever office gadget, Burke’s task was simpler. Now? It’s about confronting the razor edge where cutting-edge innovation meets potential apocalypse. How do you balance harnessing AI’s explosive power for economic growth, while staring down the barrel of job losses and ethical puzzles thicker than a Dublin fog? It’s a riddle wrapped in a mystery inside a silicon chip… and nobody’s quite cracked it yet. Feeling a bit overwhelmed? Join the club. Because if the “paperclip problem” (where a tireless AI turns the planet into scrap) isn’t enough to keep you up at night, the real trick might be convincing the workforce to dance with the very tech that could replace them. Peter Burke calls this whole mess a “very significant challenge”—a phrase that, well, sounds a bit like an understatement when we consider what’s at stake. But hey, if he’s wrong, maybe none of us will be around to argue. Intrigued? LEARN MORE

It’s not every day that you see a government minister being asked, deadpan and sincerely, about the possible extinction of humanity, writes Gavan Reilly.

Nor is it every day when the minister fielding the questions reframes the hypothetical elimination of the human race as a “very significant challenge”.

Such is the era of AI and the challenge faced by Enterprise Minister Peter Burke, who took office at a time when artificial intelligence seemed merely like a nifty office time-saver, and not the looming threat to all life as some see it.

It’s probably important to add a bit of context here. For years, AI researchers and evangelists have openly shared what they call their ‘P(doom)’ figure, mathematical shorthand for their personal expectations of an AI catastrophe that kills us all.

Three years ago, a lifetime in technological terms, a survey of AI researchers found their average P(doom) to be just under 15 per cent. One imagines that given the advances in autonomous models since then, a similar survey would turn out a higher number.

We’re also lucky enough to have deep thinkers who have already identified the philosophical issues we face in the AI rollout.

Consider the ‘paperclip problem’, in which a hypothetical factory is programmed to optimise the making of paperclips but interprets its task so literally that it treats the entire planet (including human bodies) as a resource for power and metal.

Philosopher Nick Bostrom posited this as far back as 2014: engineers have already had ten years of reminding us that AI models need to be explicitly programmed to recognise the contexts that humans take for granted.

That means that more recent revelations – that some engineers in the field have a P(doom) of greater than 10% – are not all that revelatory at all.

Moreover, those sounding the loudest alarm, like the publicised case of Jacob Coxon who quit Anthropic over precisely those fears, may not fully understand the technology themselves. Coxon thinks a renegade AI model could avoid being simply unplugged, by pre-emptively copying itself onto other computers.

That misunderstands how AI models work, it portrays them as something like a virus that infects one computer in the hope of eventually infecting another.

As Irish people have become increasingly aware, the most powerful AI models don’t simply ‘live’ on your laptop but rather in a mammoth rack of RAM in data centres. A sentient AI model would need to be especially malicious to have replicated itself into other data centres without a human realising and stepping in to stop it.

A more reasonable concern for governments like ours is how to navigate the knife-edge impact of AI on our national finances: how exactly a tech-reliant economy like Ireland can harness AI to power further growth without shedding tens of thousands of human jobs in the process.

The biggest barrier to that is probably at the very frontline: the workers most at risk of having a ‘smart’ AI model subsume their job. Those at the bottom of every professional food chain are already acutely aware of how automatable their duties are, what’s the motivation for them to start using the very tools that will eventually cost them their livelihood? Everyone knows a case where someone was unwittingly asked to train a colleague who eventually became their replacement; who’ll do it when the stakes are so high?

Even if there is no imminent threat to someone’s own job, there is a barrier in motivation. C-suite executives will be delighted to manipulate AI as much as possible, simply to maximise the bottom line. But will the worker at the bottom get a share? Is there an AI dividend coming their way?

It’s probably reasons like that which led Peter Burke to describe mass human extinction as merely a “very significant challenge” to the economy. If he turns out to be wrong, it’s not like anyone else will survive the apocalypse to tell him about it.