AI’s role in cybersecurity: Friend or foe? 

AI's threat to cybersecurity is as significant as its role in defence, making AI a double-edged sword in the battle for digital security

In the race to embrace AI many organisations are leaving themselves open to attack by forgetting the risks of cybercrime.

Yet when it comes to cybersecurity two front-runners are fighting to win supremacy.

On the one hand, are cybercriminals launching increasingly complex and sophisticated cyberattacks that are powered by AI.

On the other, are the cybersecurity teams who are harnessing AI to detect and prevent such attacks, as well as to compensate for a human talent shortfall.

The challenge is that AI is evolving so fast that organisations aren’t necessarily recognising the risks they are exposing themselves to, leaving them vulnerable. Understanding how you can win the cybersecurity race and better protect your organisation from the threat of cybercrime is vital. 

Containing the beast

AI is revolutionising cybersecurity by enabling huge advancements in detection, allowing stronger defences to be put in place. In areas such as malware detection and safe browsing AI has already proved itself, giving hope for similar advances in other areas of threat detection.

But as fast as AI drives productivity and improvements in detection its ever-advancing ability to also be at the heart of cybercrime attacks must also be recognised by organisations for its threat to be contained. 

3 categories of AI-driven crime

There are three categories of AI-driven cybercrime to be vigilant of:

1. AI-powered attacks
   The days of obvious phishing scams are disappearing fast thanks to AI-powered attacks. Large language model technology allows for sophisticated spear-phishing campaigns, while deep fake technology can mimic everything from an individual or company’s tone of voice to their style. Both can easily fool victims.
2. AI theft
   The rapid growth of AI, and the corresponding challenge in keeping up with its rollout in organisations, puts companies at risk of AI theft. Model inversion attacks skew the training data on which AI grows stronger while machine learning model theft attacks the systems at its heart.
3. AI attacks
   Similarly, criminals can attack AI directly, injecting tainted data to poison systems or sneaking harmful prompts into AI systems, for example.  

Three ways to use AI to boost human efficiency in cybersecurity

Adopting AI to combat existing and future threats to cybersecurity, particularly those powered by AI, is crucial. This is particularly important as the need for programming skills to launch cybercrime diminishes with the adoption of AI, widening the pool of potential cybercriminals and therefore also the risk of attack.

As well as better enabling detection using AI as part of your cybersecurity weaponry can also help overcome skills shortages or to allow human effort to be directed elsewhere. This allows you to combine both AI and human expertise within your cybersecurity defence strategy.

There are three ways of doing this effectively:

1. Strong AI governance
   This allows you to maintain control and requires a mix of clear guidance over AI model use and commitments to effective data management to reduce risk.
2. Awareness and training
   Improved awareness and training of AI cybersecurity risks will help to empower your teams to recognise AI-fuelled attacks and to detect and halt them. But such training must be a continual process. There’s no time in this race to stand still.
3. Protection
   Dynamic protection of your systems and intellectual property should put AI at the heart of product development to spot vulnerabilities before they are exploited. Improved threat intelligence and stronger infrastructures also help to fight off threats.