AI Cybersecurity

How I Learned to Stop Worrying and Love the Kill Switch!

This is AI blog number four, and by now I thought I’d be off down the rabbit hole exploring the deep philosophical implications of AI. You know, the big stuff—ethics, consciousness, the meaning of intelligence. But no. Instead, I’ve just completed a cybersecurity course and I’m still recovering from the realisation that I was far more vulnerable than I ever imagined. Frankly, I could tip right into full-blown paranoia.

The irony? So many of these shiny new AI tools we’re being sold are designed to help us stay compliant—tick the regulatory boxes, meet legal standards, keep data protected. Helpful, yes. But here’s the catch: the very tools meant to protect us may themselves be wide open to the kind of cyber-piracy they claim to prevent. It’s like fitting a burglar alarm and then leaving the back door ajar.

I now see cybersecurity like one of those battered treasure chests in Pirates of the Caribbean—brimming with gold coins. But in our world, those coins are data: your data, my data, everyone’s data. And the chest? It’s got a dodgy lock that can be prised open by any reasonably competent hacker armed with malware, ransomware or deepfake tools. AI, paired with biometric and synthetic tech, makes it even easier—because now the pirates don’t just steal the gold. They can look like you while doing it.

The reality is grim. They don’t need a reason to hack you. It might be financial. It might be political. It might just be chaos for chaos’ sake. You don’t always get the motive; you just get the mayhem. A cyber-attacker quietly slips into your system (usually after someone innocently clicks on a dodgy link or gets stuck in a phishing scam) and they lock onto your data, passwords etc.  They are brazen; recently scammers send a false email from me asking the Finance Dept to agree a set of vouchers that had they scanned the codes, we would have funded a scammer spending spree.

Its not just us who are at risk, big institutions aren’t any safer. Remember WannaCry? That 2023 cyberattack paralysed the NHS, among others. It locked systems, froze appointments, shut down operations. WannaCry was based on tech developed by the US National Security Agency but it still got weaponised. The scale of it was terrifying—and we only narrowly avoided total disaster thanks to a lone analyst who stumbled across a kill switch in the code and shut it down. It cost him £10 to buy the unregistered domain and pull the plug. £10!

We rush headlong into AI with breathless optimism, expecting it to sort our admin, speed up decision-making, and make everything more “efficient”. In truth, we often spend more time trying to log in, connect to the right platform, or decipher robotic helpdesk messages written in technobabble. And when all else fails, what do we want? A phone number. A human. That elusive thing called trust.

And yet, trust is in short supply. Why? Because despite the branding, AI isn’t artificial in the way people think. It’s designed by humans, trained on human data and full of human bias. It doesn’t think. It doesn’t understand. It doesn’t “know”. It simply predicts.

Let’s bust a few myths:

• AI isn’t artificial: It’s built by teams of human engineers and run on physical infrastructure—data centres, chips, electricity. Nothing intangible about it.
• AI isn’t human: It doesn’t “think”. It doesn’t have intention. It completes tasks it’s been trained on.
• AI isn’t always right: It “hallucinates”, fabricates, gets things wildly wrong—and does so confidently.
• AI isn’t neutral: It inherits bias from its creators and the data it’s trained on. Garbage in, garbage out—just faster and harder to trace.

So, where does this leave us?

The Six Big Ethical Questions

1. Privacy and Data Use: Most of us are unwittingly handing over personal information to AI systems—think Copilot, ChatGPT, smart watches, loyalty cards—without reading the small print. Often, you have to actively opt out of being included in model training. How many of us bother?
2. Consent: Is it ethical to input someone’s data into an AI system—say, to summarise meeting notes or write a letter—without their knowledge or permission?
3. Transparency: Much of AI is still a black box. You see what goes in and what comes out, but not how the decision is made. That’s dangerous in sectors like finance, healthcare and criminal justice.
4. Bias: AI reflects and reinforces systemic bias. If the data is skewed, the outputs will be too.
5. Accountability: Who’s responsible when AI gets it wrong? You? The designer? The machine?
6. Regulation: Countries are scrambling to create laws and frameworks. The UK wants to be a leader, but right now, adoption is piecemeal. The government’s AI Opportunities Action Plan talks up potential—economic prosperity, smarter public services, data-driven innovation. But even the National Audit Office concluded that adoption across the public sector is still at an early stage. Too many pilots, not enough joined-up thinking.

But like it or not, AI is already in our homes, our settings, our GP surgeries. Your phone wakes you up, your watch tracks your heart rate, your supermarket logs your snack choices, and your binge-watch habits are carefully recorded and fed into an algorithm somewhere. Facial recognition, license plate tracking, and AI-powered cameras can collectively paint an eerily detailed picture of your entire day.

It’s not sci-fi – it’s now. And while some of it is helpful, too much of it is opaque, unaccountable, and potentially dangerous—especially when combined with lax cybersecurity and vast, unregulated data collection.

We are sleepwalking into the future!

At the very least, we should be walking with our eyes wide open. Therefore, take a cyber security course. You need to know the basics. Make an action plan to review what software your team is using, who has access, what does data compliance mean for you and your setting and learn how to opt out of cookies. In a world where a £10 domain name can stop an international cyberattack, imagine what a bit of knowledge in your hands could do.