The Interview | AI-powered cyberattacks: Defend your business

And obviously the cyber landscape is already very complex.

How is AI changing that?

Well, AI is changing everything, as we know, in everyday tasks.

You can't go into a shop or a pub or a restaurant and not overhear someone talking about AI.

For us, AI lowers the amount of time and specialist effort needed to carry out many of our tasks.

It's a positive for us as a defender, but attackers also use the same capability.

And there's some major companies that listeners will have heard of, Anthropic and OpenAI.

OpenAI in particular published a letter at the start of this month which warned that AI-enabled attacks are expected to become much more widespread and sophisticated.

while long-standing weaknesses such as unpatched software, excessive access and weak sign-in methods have left organisations exposed.

Their open letter started with the headline sentence, which is pretty impactful, that said we have a limited time to strengthen cyber defences.

And the letter talks about months rather than years.

And that's the kind of timeline that we're seeing generally develop for everything related to AI.

So whilst some might point to OpenAI and Anthropic as the leading threat enablers, they're not alone in providing the technology capability.

Similarly, Anthropic published a lengthy risk assessment for its mythos model, and listeners will have probably heard about some of that.

The threats are not solely limited to cybersecurity and their risk assessment.

That's the bit that grabbed the headlines.

Interestingly, chemical and biological weapons risk, you know, can AI create a novel chemical weapon, occupied several tens of pages as a risk assessment matter.

So the important point for a non-technical audience is the speed and scale that AI development is happening at.

For an attacker, they can research targets, create convincing messages, rewrite them for different audiences, and repeat the process far more quickly.

And we expect as a company and cyber security professionals generally expect that a lot of this kind of attack chain, as it's called, is going to be automated or is automated.

It doesn't make every attack successful, but it does narrow the time a business has to respond.

And it means that a business has to sharpen their opportunity to respond.

The letter from OpenAI stressed a positive element to its warning.

And interestingly, obviously, the warning is in the context of a company that's making the machine that's enabling, potentially enabling the types of attack They stressed a defender's window, fairly limited, not too long in time, and they suggest that organisations can use AI-supported tools to find and fix weaknesses, wider monitoring and sharpen proven fixes.

The right response for us, for a business, is therefore a focus on timely improvement rather than a sort of blind panic approach.

And obviously the cyber landscape is already very complex.

How is AI changing that?

Well, AI is changing everything, as we know, in everyday tasks.

You can't go into a shop or a pub or a restaurant and not overhear someone talking about AI.

For us, AI lowers the amount of time and specialist effort needed to carry out many of our tasks.

It's a positive for us as a defender, but attackers also use the same capability.

And there's some major companies that listeners will have heard of, Anthropic and OpenAI.

OpenAI in particular published a letter at the start of this month which warned that AI-enabled attacks are expected to become much more widespread and sophisticated.

while long-standing weaknesses such as unpatched software, excessive access and weak sign-in methods have left organisations exposed.

Their open letter started with the headline sentence, which is pretty impactful, that said we have a limited time to strengthen cyber defences.

And the letter talks about months rather than years.

And that's the kind of timeline that we're seeing generally develop for everything related to AI.

So whilst some might point to OpenAI and Anthropic as the leading threat enablers, they're not alone in providing the technology capability.

Similarly, Anthropic published a lengthy risk assessment for its mythos model, and listeners will have probably heard about some of that.

The threats are not solely limited to cybersecurity and their risk assessment.

That's the bit that grabbed the headlines.

Interestingly, chemical and biological weapons risk, you know, can AI create a novel chemical weapon, occupied several tens of pages as a risk assessment matter.

So the important point for a non-technical audience is the speed and scale that AI development is happening at.

For an attacker, they can research targets, create convincing messages, rewrite them for different audiences, and repeat the process far more quickly.

And we expect as a company and cyber security professionals generally expect that a lot of this kind of attack chain, as it's called, is going to be automated or is automated.

It doesn't make every attack successful, but it does narrow the time a business has to respond.

And it means that a business has to sharpen their opportunity to respond.

The letter from OpenAI stressed a positive element to its warning.

And interestingly, obviously, the warning is in the context of a company that's making the machine that's enabling, potentially enabling the types of attack They stressed a defender's window, fairly limited, not too long in time, and they suggest that organisations can use AI-supported tools to find and fix weaknesses, wider monitoring and sharpen proven fixes.

The right response for us, for a business, is therefore a focus on timely improvement rather than a sort of blind panic approach.