1. Podcasts
  2. This Is Why
  3. Why web attacks show we’re all vulnerable
This Is Why

22 Oct 2026 · 14m · 7 chapters

Episode artwork

Why web attacks show we’re all vulnerable

In short

A severe PHP (widely used web back-end language) vulnerability shows how AI can both uncover and accelerate web attacks, leaving many websites vulnerable; the episode stresses the gap between vulnerability and exploitation and what the public should do.

Guests

Roland Manthorpe, Sky News technology correspondent. Gareth Barlow is the host.

Key claims

PHP is used on about 70% of sites; the flaw was coded ~14 years ago and was discovered recently with AI. Cloudflare reported millions of exploitation attempts. No major PHP-based hacks are reported yet, but unsupported sites may be exploited as staging posts, potentially leading to later credit-card theft. Cybercriminals are the likely exploiters; attack volume may be deceptive.

Notable examples

public bodies temporarily taking down UK sites (Office for Budget Responsibility, NHS Scotland, Oxford County Council); prior hacks mentioned (ASOS push alert; French supermarket refrigerator tampering).

Written by AI. May contain mistakes.

Chapters

1:10

The PHP Vulnerability Crisis

Exploring the widespread PHP vulnerability affecting millions of websites.

“including giants such as Amazon and Wikipedia.”
1:46

Understanding PHP and Its Importance

Explanation of what PHP is and its critical role in web functionality.

“so luckily we're joined by our technology correspondent, Roland Manthorpe.”
3:38

AI's Role in Finding Vulnerabilities

Discussion on how AI has enhanced the discovery and exploitation of web vulnerabilities.

“And the whole backstory of exactly who discovered it is kind of complicated.”
5:07

The Nature of Cybercriminals

Examination of who is attempting to exploit the PHP vulnerability and their methods.

“But so far, no one has reported a major hack based on PHP, which is good news.”
7:34

Current Cyber Threat Landscape

Analyzing whether the frequency of cyber attacks has increased due to new vulnerabilities.

“Who's going to be looking to try and leverage this?”
10:38

The Future of Cybersecurity with AI

Speculating on how AI could both help and hinder cybersecurity in the future.

“you can watch your dog or your pet you name it everything can be connected to the internet are we at risk of being too connected are we opening ourselves and our lives up to too much potential exploitation?”
12:00

Advice for the Public on Cyber Safety

Guidance on what individuals should do in light of the PHP vulnerability.

“And in the short term, it does stand to reason, I think, that it should benefit attackers over defenders.”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On This Is Why, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:01Sky News, the full story first.

0:11The rise of AI means our personal data might never be safe again. And this is why.

0:35It gives me glucose insights without finger sticks. Learn more at freestyllibre.us. Finger sticks are required if your glucose alarms and readings do not match symptoms, or when you see check blood glucose symbol for the first 12 hours. The views expressed should not be used for medical diagnosis or treatment as a substitute for professional medical advice. Individual circumstances may vary for prescription only. Safety info found at freestyllibre.us.

0:57If you think you've had a tough day at work, spare a thought for the IT departments of governments and businesses around the world. A cyber security flaw in something called PHP has left millions of websites vulnerable to attack, including giants such as Amazon and Wikipedia. In the UK, the Office for Budget Responsibility, NHS Scotland and Oxford County Council are among the public bodies that decided to temporarily take down their sites. That's because, never once to miss an opportunity to steal our personal data, hackers are responding in a way their name suggests. The head of one major tech company claims there have been millions of exploitation attempts against the PHP flaw since it emerged.

1:36The vulnerability was identified with the help of an AI agent, but artificial intelligence has also made it easier and quicker for hackers to exploit it. I'm Gareth Barlow, and it's a pretty complicated story, so luckily we're joined by our technology correspondent, Roland Manthorpe. And Roland, the first question surely has to be, what is PHP, why are we bothered, and what's the challenge here? Okay, well, we're talking about it because of AI, and I'll explain that. But first, maybe it's just if I might help if I explain what PHP is. PHP is a programming language that is used incredibly widely across the internet, by some estimates on 70 % of sites.

2:14And it's a back-end thing. But you probably will have interacted with it in the last few days if basically you've used a website for anything. Because you know if you say, for instance, if you were buying something on some website and you put it in your basket and then you went to the basket and all the details were there. How does it get there, right? Because that's a completely individual page. You see that and it's completely different to anything else anyone has seen. That is PHP. PHP is reaching out to the servers and putting those details in the website. And you can imagine, right, this is an incredibly useful, important technology.

2:51It's incredibly widely used. And this is why news of a vulnerability, a severe vulnerability in PHP has hit so hard. because this vulnerability that's been revealed will allow people to add their own instructions into websites without logging in. And you can imagine, right, in this case, you are really in the back office with the keys. You can pretty much do whatever you want on a website. So it's a combination of something extremely widely used and, if exploited, something really bad. The timescale of this is all really interesting, isn't it? Because actually it turns out that the The kind of error was coded in years ago or was it developed years ago, but we didn't know about it for years.

3:32And then we kind of knew about it. And then AI really cracked the case in recent weeks. They've said that this error was coded 14 years ago and it's only just been discovered by AI. And the whole backstory of exactly who discovered it is kind of complicated. But essentially different groups of researchers using AI found this vulnerability. and this is really an example, a good example of what is happening right across the web. AI has supercharged our ability to find vulnerabilities. People put out charts of this kind of stuff and it's like one of those charts which kind of tickles along for a while, you know, for a long time, just normally, and then suddenly AI comes along and it goes vroom.

4:11And most often these vulnerabilities are, yeah, they've been hidden there for ages, they've been lurking away because all code is vulnerable. All code has bugs. And suddenly were able to find them. Before, we thought we were fine. We knew we weren't, but we knew no one could probably find this stuff. Now we can. And yeah, that creates a whole lot of problems. Is it creating problems with this PHP bug, glitch, error, or has nobody yet managed to exploit it? Yeah, well, that's a really good question. I think it gets to a core distinction that is sometimes kind of missed in these things. There's a massive difference between a vulnerability and an exploitation.

4:47You know, like you might know that you've left the attic window open, But as long as no one climbs in, then you're okay. Right now, we've got this very important vulnerability, and there's a race to patch it before attackers get in. They are trying. Cloudflare, a company that protects websites, has already reported millions of attempts using this vulnerability. So we know they're trying. But so far, no one has reported a major hack based on PHP, which is good news. I think basically what we can say is that they're inevitably, inevitably will be hacked based on PHP because there are lots of old, unsupported websites that can't even be patched.

5:29Attackers will be able to gain access to those. They're not that important, obviously, because they're often old and abandoned. They're not that major, but they can use those for staging posts, for other attacks. The attackers get in, they do something like change the code ever so slightly so that if you enter your credit card, it sends them all the details. And that is how months later, we find out that a whole lot of people have money stolen, and it all goes back to this PHP. It feels as though this is a bit of a two faced AI story, right? Because AI in one sense has done some great work, it's found a vulnerability that could affect 70 % of websites around the world is astronomical figure.

6:07At the same time, AI could also be used, we're always hearing, to then exploit these vulnerabilities. So how do we find the balance here between utilising the capacity and then also defending against the capacity? Oh, yeah. I mean, that's the ultimate question. That's the ultimate question, right? Figure that out. You can retire early. Yeah, yeah, yeah, exactly. Although, can I just say that maybe before you think about how this stuff is repaired or attacked, it's important to think about the communication around this, because actually, that is one of the most difficult bits. Because I mean, it's kind of weird, isn't it?

6:40That basically, we've found this really important vulnerability. And the only way that we have of telling everyone about it is just to put it on loudspeaker so that everyone, even the attackers, know about it. There's a real communication difficulty in getting the message out in a way that doesn't make everyone vulnerable. But anyway, basically, this is the situation now. Attackers have AI, defenders have AI, and everyone's in a race to use it. I mean, whether people are actually using AI as much as you might think is still an open question. But going forward, this is going to be the situation.

7:13We've got the same tools and we're all racing to use them. And in a way, that is just the history of cyber. Going back to that figure of millions of attempts to try and exploit this vulnerability, I have the Gareth Barlow scale of hackers. I have pesky kids at one end and I have, oh God, the Russians at the other. Between that kind of, that optic of who it might be trying to do this, is it just people that are idly curious? Is it cyber criminals? Is it nation states? Is it rogue actors? Who's going to be looking to try and leverage this? Cybercriminals, I'd say. I mean, I've got to say pesky kids can cause some of the biggest damage.

7:45There was quite and quite pesky kids that brought Jaguar Land Rover down. But yeah, it's cybercriminals. And that metric of attempts can in some way be deceptive. Because remember, it's computers doing this, running over and over and over again, trying again in lots and lots of different ways. So we don't even know really how many people are behind this. But the kind of people who would be looking to exploit this essentially are cybercriminals. In the last couple of weeks, we've talked about ASOS being hacked and that push alert on people's phones. We've talked about people gaining control of French supermarket refrigerators and whacking the temperature up so that meat's spoiled and, well, ultimately some of them set on fire.

8:26Now we're talking about a vulnerability in 70 % of the world's websites. Are we seeing more issues or are we just seeing better reporting and understanding of issues that maybe we're already about, but we just talk about them more than we did? The data on this isn't great, but so far, somewhat to my surprise, I would say we are not seeing more issues. We are seeing roughly the same amount, but we're now associating them with this looming fear that this might be the start of something really bad because of AI. And it's quite perplexing in a way. I've been expecting there to be a massive cyber wave that might suddenly, let's say, double or maybe even times by 10 the amount of cyber attacks that we currently get.

9:13And by the way, for reference, it's estimated that cyber attacks currently cost the UK 0.5 % of GDP. I mean, double that or times that by 10 and you're looking at something extremely major. But so far, we haven't. When I've puzzled over this with people in the field who work on this day in, day out, often what they say is, well, it seems to us like the guys who do this actually can't be bothered to upgrade yet. Business is good. It's quite expensive to run AI. They'd have to, in this ultra-sophisticated way, they'd have to learn a whole new set of skills, buy some servers. Might cost a ton of money.

9:53Can't be bothered. We definitely see that they're using it for things like customer service and sort of their HR functions and stuff like that. Because, I mean, these cyber gangs, they're now kind of like large organisations in a way, but they don't seem to be using it, say, for ransomware. And, for instance, ransomware was bad before AI got really good at cyber. It's bad now. But it hasn't really changed. Of course, the worry is that it will in the future, but so far we haven't seen it. websites are being patched our mobile phone providers are always providing updates you can get software to protect your laptop but now in 2026 your computer's not just connected to the internet but your doorbell is your smart meter your fridge the camera that is in a ball that you can watch your dog or your pet you name it everything can be connected to the internet are we at risk of being too connected are we opening ourselves and our lives up to too much potential exploitation?

10:51I mean, maybe, but it's probably too late. Yeah, I'm afraid to say, I don't think that's going to change. I guess if you look forward, there's a real possibility that we might actually not end, but really dramatically improve our situation thanks to AI. We have never even been able to come close to writing code that doesn't have bugs and vulnerabilities. AI could do that for us. And in theory, it should benefit the defenders over the attackers in the long term. And I look forward to some kind of scenario where you have sophisticated AI systems with tens of thousands of agents running over code automatically, detecting vulnerabilities as soon as they emerge, fighting off attackers.

11:43and that actually that makes us more secure and we're really able to enjoy the benefits of all this connected stuff without this constant threat of attacking. I feel like that should be possible. But of course, you know, sort of the long term could be quite a long way away. And in the short term, it does stand to reason, I think, that it should benefit attackers over defenders. So we might be in for a bumpy few years. Bringing everything back to today, for any member of the public, anybody. What does PHP mean for them today? Do they need to go and change passwords? Do they need to get off the internet for a bit?

12:22What's going to happen to us, right, as a result of all of this? Well, great news. Nothing. We don't need to worry about it beyond the things that we normally do. Because just visiting a website with this PHP vulnerability doesn't infect you. Our phones, our laptops, any of our devices, they haven't been changed. Really, we just face the same problems that we did before, but to a slightly greater extent, in that you might go onto some website, might be an older website, it's been hacked, and then you're in trouble. But even then, often, ultimately, what you're looking at is the same stuff we know already.

13:02The attackers have managed to get some of your data, and then they use that to try and to email you and execute a phishing attack on you. And there are, you know, say like email systems, they have a lot of defenses to prevent that. And also there is this gap, right? There's this gap between vulnerability and exploitation. And sometimes that gap is just us and our knowledge and our ability to spot these things coming in. Unfortunately, it's always the case with this stuff, and it's been so well reported, is that it's more vulnerable people, especially older people who often get hit with this kind of stuff.

13:33And if you do, it can be really bad. But for most people, basically, the advice is, look, just continue with your normal cybersecurity, use two-factor authentication, use pass keys, look out for dangers. But also, you know what, carry on with your ordinary life doing everything you do online because these systems are, I mean, really, the miracle is that they stay up at all and that they're as secure as they are. Our core systems haven't as yet, so far as we can see, been affected. that was Roland Manthorpe good to have you on the podcast good to have you listening to the podcast dear audience that's it for today message the show send us an email why at sky.uk this is why we'll be back tomorrow

From the publisher

PHP. The three little letters few people realised were so vital for most of the websites in the world.

But a vulnerability in a piece of programming virtually all of us had never heard of has affected everyone from small businesses to major governments.

Millions of cyberattacks have followed in the wake of the problem being made public. But the fact is, your personal information – from email addresses to bank details – was already at risk.

AI is increasingly both poacher and gamekeeper – discovering the vulnerability in the first place but also playing a vital role in cybersecurity. So, is it time to rethink our online lives?

Gareth Barlow speaks to Sky’s technology correspondent Rowland Manthorpe.

Have you got a question for the show? Email us: [email protected]

More from This Is Why

All episodes