Kenyan businesses are moving from using artificial intelligence (AI) for simple tasks like answering customer questions to deploying systems that can execute tasks on their behalf, and experts point to an emerging cybersecurity challenge – how to protect the digital identities and privileges granted to these AI agents.
Conventional chatbots such as ChatGPT, Gemini and Claude mainly generate responses. But AI agents can be connected to company applications, databases and internal systems to perform tasks with limited human intervention.
This makes them useful for automating business processes, but also creates another potential entry point for cybercriminals if their access is compromised or poorly controlled.
For Kenyan businesses adopting AI to streamline operations, an agent with extensive permissions could expose sensitive information or allow an attacker to move deeper into corporate systems if its identity is hijacked.
“AI agents are what we call privileged users in your environment,” Robert Ng’etich, the team lead for threat intelligence in East Africa at the global IT services firm NTT Data, told the Business Daily.
“For such users, one of the biggest risks they present is that if their identities are compromised, they can lead to a broader business compromise.”
The risk reflects a wider shift in cybersecurity, where attackers increasingly exploit existing identities and login credentials rather than relying only on technical vulnerabilities to break into systems.
Using legitimate credentials can help hackers disguise their actions as routine user activity, making it harder for organisations to identify an intrusion.
Because AI agents gain access to business applications and data, cyber experts say companies will need to extend identity protection and access controls to these non-human users, just as they do for employees and other privileged accounts.
This includes ensuring that agents have access only to the information and systems needed for their assigned tasks, monitoring their activity and protecting the credentials that allow them to operate. An agent that can access several business systems may become an attractive target if its permissions are not carefully managed.
The threat is emerging as cybercriminals increasingly harness AI to improve the speed, scale and credibility of their attacks.
The Communications Authority of Kenya (CA) detected 11.1 billion cyber-threat incidents in the year to June 2026, a 29 percent increase from 8.6 billion a year earlier.
Distributed denial-of-service attacks, where hackers flood websites, servers and networks with huge volumes of malicious traffic to make them inaccessible, rose 114.3 percent to 72.16 million in the year ended June 2026, up from 33.68 million a year earlier.
This was the highest growth among the cyber threats tracked by the regulator, highlighting the growing exposure of businesses and public systems as more services move online.
Web application attacks targeting flaws, poor coding, or security gaps in websites, web services, and application programme interfaces increased 99 percent to 51.51 million, from 25.89 million in the previous year.
Cyber experts say attackers are using AI to generate more convincing phishing messages, translate scams into local languages such as Swahili, and create deepfake voices and videos that imitate real people.
Read: How AI misuse is fanning digital crime in businesses
These techniques can be used to trick employees into disclosing information, authorising payments or carrying out instructions they believe came from a colleague or senior executive.
“Before, it used to take a lot of time for attackers to find ways to exploit systems. But now, with the help of AI, what would have taken maybe three weeks to one month is now taking hours,” said Mr Ng’etich.
“We are seeing a lot more attacks on websites and other systems because attackers are using AI to enhance their vulnerability research, increasing the speed with which they are finding loopholes.”
Still, the same technology is being adopted by cybersecurity teams to strengthen their defences, setting up a contest in which both sides are seeking to improve their speed and effectiveness.
Security operations centres (SOCs), which monitor networks and investigate suspicious activity, are using AI to analyse and correlate far more signals than human analysts could reasonably examine manually.
Events such as a user login, access to a file and other seemingly minor activities can be assessed together to reveal patterns that might indicate a wider attack.
Mr Ng’etich says AI can help security teams examine hundreds of data points and connect events that, viewed individually, may appear insignificant.
“We can also do summaries using AI, report on incidents using AI, whether to technical or executive teams in a business,” he said.
“The SOC is also able to use AI for vulnerability search too, the same way as attackers. And because of that, they can arrest, or they can address vulnerabilities at a much faster pace.”
Security practitioners are also deploying AI agents to handle repetitive work, including case summaries, event correlation, prioritisation and certain standardised responses, such as blocking elements associated with an attack.
The aim, he says, is not to replace human analysts but to help them respond more quickly and concentrate on decisions that require judgement. It increases the volume of security events they investigate and shortens the time between detecting a threat and taking action.
Follow our WhatsApp channel for the latest business and markets updates.
Click Here For The Original Source
——————————————————–
