“AI itself is neither inherently good nor bad – but it has the potential to be both. In the hands of defenders, it can offer powerful protection from cyberattacks; in the hands of cybercriminals, it can be used to exploit human vulnerabilities on a massive scale. It all depends on who is behind the keyboard,” – Allan Juma, Cyber Security Engineer at ESET East Africa.
Artificial intelligence (AI) is transforming cybersecurity – for better and for worse. Cybercriminals are leveraging AI advances to coordinate more sophisticated attacks, while security teams deploy improved defence mechanisms, using the same technology. For businesses across Africa, who have prioritised digitisation over the past decade, this creates new vulnerabilities that need to be addressed, not just at a tech level – but a human one.
“Advanced social engineering is one of the top AI-enhanced cyber threats on our radar. Generative AI and large language models – ChatGPT, for example – make it easier for cybercriminals to craft convincing phishing emails that imitate executives or colleagues really well. These models are also skilled with translation, allowing attackers to target new regions and smaller countries with niche dialects,” says Juma.
Using AI, attackers can automate vulnerability scanning – increasing the speed and scale at which they exploit weaknesses in a business’s security. This includes identifying compromised internal accounts, which they in turn use to send phishing emails or deepfake impersonations (audio or video generated by AI to depict real people, including managers, CEOs, and finance officers). As the technology improves, it becomes increasingly difficult for employees to discern legitimate communications from fraudulent ones.
“This is where the human element comes in. A significant percentage of data breaches are the result of human error. Cybersecurity awareness training is essential to combatting AI-driven cybercrime. A lack of knowledge is the greatest threat to cybersecurity in any business – and cybercriminals know that, and they’re going to take advantage of it,” says Juma. Research conducted by the Google Threat Intelligence Group (GTIG) found that cybercriminals are using Google’s AI model Gemini for research and content generation. According to GTIG’s Adversarial Misuse of Generative AI report, released in early 2025, this includes developing target personas and messaging, as well as translation, localisation and finding ways to increase their reach.
Where training and vigilance fail, AI driven defence mechanisms may be the key to protecting business operations from cyberthreats. Security teams can use AI to analyse patterns and predict cyber threats before they happen – as well as automate responses to detected threats, reducing reaction times and mitigating potential damages more effectively.
“Actually, AI has been part of cybersecurity approaches and software for a long time – way before it became a big talking point for the general public. This gives us something of an advantage, because we’ve been able to spend more time with it and integrate it into our systems,” says Juma. “The danger with AI becoming such a pervasive topic is that we have forgotten how dangerous it can be. Businesses need to keep it top of mind and be aware of the risks – so that they can address them.”
