The cybersecurity landscape is rapidly changing, with hackers increasingly utilising Artificial Intelligence (AI) to carry out sophisticated attacks on their targets. To defend against these AI-driven threats, organisations must adopt data protection solutions that harness AI – essentially, “fighting fire with fire”.
The IT space today is a battlefield, with cyberattacks getting smarter, faster and harder to detect. Brute force attacks, where millions of leaked passwords are tried until one works, are still common, but AI is changing the game. Instead of blindly guessing, attackers now use AI to analyse systems, identify weak points and strike with precision.
Phishing, however, remains the easiest way in, as all it takes is one click from a valid user. But with AI’s mastery of language, phishing emails have evolved from clumsy, error-filled messages to polished, native-level English, making them far more convincing.
Organisations must realise that security is not a single wall; it consists of layers, like an onion. There is no silver bullet, and hackers look for the kink in the armour. This ‘armour’, so to speak, must include moats, watchtowers, drawbridges and reactive defences.
Layered approach for applying AI
The same layered approach applies to AI threats. First, educate your teams, as AI is used by both defenders and attackers. If users are not aware that AI can craft flawless phishing emails, they will not spot the danger. Spelling mistakes used to be red flags, but now phishing reads like polished British English, making clicks more likely.
So, awareness is step one, followed by smart boundaries and AI-powered tools that detect behaviour, not just signatures. Traditional platforms must evolve, and many now include AI, however, adoption must be strategic. Merely incorporating AI is not enough. It needs to be integrated with a purpose to counter increasingly polymorphic threats.
Polymorphic malware is a game-changer. While traditional antivirus tools rely on signature databases to detect threats, polymorphic attacks constantly change form, making those databases ineffective. These are zero-day threats, unseen, unpredictable and hard to trace.
To counter this, organisations must integrate AI into their security stack. Behavioural analysis is key as AI can detect unusual patterns, like an endpoint suddenly scanning for credentials or a server moving massive files unexpectedly. These anomalies often flag real threats, but sometimes, it is just a human doing routine work, like closing financial books.
Machine learning helps, but it is still maturing, and it needs context and careful calibration. Hence, simply bolting on AI is not enough; its adoption must be strategic to keep pace with increasingly sophisticated attacks.
Using AI to prevent alert fatigue
At the same time, organisations should be aware that alert fatigue is a growing issue in IT. With multiple tools monitoring networks, storage and endpoints, security teams are overwhelmed by constant alerts; many of which are false positives. This means analysts must sift through thousands of logs to find real threats, which is not sustainable.
AI-driven Security Information and Event Management (SIEM) tools help by correlating and filtering alerts. They recognise patterns, like a failed login followed by successful multifactor authentication, and dismiss noise. This shifts security from reactive to intelligent, reducing manual workload.
Machine learning lays the foundation, but agentic AI takes it further by making decisions without human input. Adopting this level of AI requires thoughtful integration and is a strategic evolution for any organisation serious about cybersecurity.
AI is also pivotal in ensuring data compliance and safeguarding sensitive information in complex digital environments, especially with regulations like POPIA, ISO 27001 and DORA. As businesses generate more data, through cloud apps, virtual machines and rapid deployments, data sprawl becomes a compliance nightmare.
Organising chaos with AI tools
AI helps by organising chaos, as it can scan files for personal identifiers, assess access permissions, and flag risks, like a CEO’s contract being readable by everyone. Agentic AI goes further – not just alerting but acting – locking down files and enforcing compliance automatically. It can categorise, secure and align data with regulatory standards. It is not just helpful but essential for managing risk in today’s data-heavy environments.
AI can be described as a “bolt-on”, like adding a supercharger to a car. It can make your business faster, smarter and more compliant, but only if integrated thoughtfully. AI must align with the systems already in place, without causing disruption.
That is why human training and awareness are key, as AI needs direction to be effective. Ultimately, adopting AI is not just a tech upgrade but a strategic journey that requires both human and operational alignment.
By Hemant Harie, Group CTO at Data Management Professionals South Africa
