Skip to main content
William Wright

William Wright

William Wright is the CEO and founder of Closed Door Security, a Scottish cybersecurity consultancy, and Scotland's first Chartered Cyber Security Professional specializing in penetration testing and red teaming.

Nov 11, 2026

1:39
Now, where do you land in AI coming into this? Does it change necessarily the calculus to make generalists more capable or are now specialists essential? Any thoughts there?
1:54
Yeah.
1:54
So I think there's really two trains of thoughts here.
1:57
So I'm a firm believer that a generalist is much more powerful than a specialist in 95% of scenarios.
2:06
Somebody who's extremely good at web apps can maybe open the door into an organization.
2:12
But without a generalist there to be able to cover all the stuff that comes after that, they sometimes get left high and dry.
2:17
Now, specialists absolutely have their place.
6:36
So William, where do you come out on that? Is attacking AI a whole new branch of pen testing? Or is it different from other, just integrating into your existing pen tests?
1:39
Now, where do you land in AI coming into this? Does it change necessarily the calculus to make generalists more capable or are now specialists essential? Any thoughts there?
1:54
Yeah.
1:54
So I think there's really two trains of thoughts here.
1:57
So I'm a firm believer that a generalist is much more powerful than a specialist in 95% of scenarios.
2:06
Somebody who's extremely good at web apps can maybe open the door into an organization.
2:12
But without a generalist there to be able to cover all the stuff that comes after that, they sometimes get left high and dry.
2:17
Now, specialists absolutely have their place.
6:36
So William, where do you come out on that? Is attacking AI a whole new branch of pen testing? Or is it different from other, just integrating into your existing pen tests?
54:24
So over those years, what stood out? Let's talk through one of these engagements that you've
54:30
been on.
54:30
Oh, so there's a lot, as you can imagine.
54:34
The most prominent one that I can talk about, I can't say who it was, but I can talk about it, was there was a bank, a UK bank.
54:42
They had asked us to just basically, whatever you can do, go for it.
54:47
Here's a bank of money.
54:48
That's how much you've got.

10 MINS LATER

64:23
so as we as we come to a uh you know up to time here i wanted to take a little a little bit here to give you opportunity and and to hear a little bit more about closed-door security, the stuff that you guys are doing, how to reach you, those types of things so that folks that are interested in having you steal their yachts or similar know how to reach you and really what your focus area is and anything like that that would be beneficial to you.
6:30
To double down on the ThreatLad pentesting, is it fair to say that initial assessment can completely derail your work? So if the initial assumptions are even slightly wrong, it completely derail the results of your, you know, work down the line?
6:59
Yeah.
7:00
And because there's a couple of different outcomes expected from a threat-led pen testing.
7:05
One of them is the detection and response.
7:08
So in order to have accurate detection and response, you need to be running accurate TTPs against whatever the infrastructure is.
7:14
So if you make assumptions, you know, so when we are initially scoping out, obviously we don't have threat intelligence at the start.
7:21
So we In fact, I can state this, we don't do threat intelligence.

7 MINS LATER

speaker_1NARRATOR
14:13
Turns
29:43
So it moves to the next point and an area of your expertise, human engineering.
29:49
Yeah.
29:50
So the interesting thing about cybersecurity as an industry is approximately one year ago, there was this hype about how humans and the entire cybersecurity industry can be replaced by AI.
30:06
myself and a number of other people working on AI working bodies and stuff with various regulators go through it and actually speak to companies.
30:14
What are you using it for? How are you developing it? What's the architecture? What's the actual use case of this? And the vast majority of cybersecurity companies did exactly that.
30:24
They went, Everything in AI, oh no, this isn't working.
30:28
Let's call that back a bit.

14 MINS LATER

44:25
I mean, that's kind of

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.