Kenyan businesses are moving from simple AI tools to AI agents that can execute tasks, creating a new cybersecurity challenge around protecting the digital identities and privileges of these agents.
Unlike chatbots such as ChatGPT, Gemini and Claude that only generate responses, AI agents can be linked to company applications, databases and internal systems to act with limited human oversight.
This capability makes them useful for automating processes but also creates a potential entry point for cybercriminals if the agents’ access is compromised or poorly managed.
Robert Ng’etich, team lead for threat intelligence in East Africa at global IT services firm NTT Data, said AI agents are “privileged users in your environment” and warned that compromised identities could lead to broader business compromise.
Experts note that attackers increasingly exploit existing identities and login credentials rather than relying solely on technical flaws, allowing them to blend into routine user activity.
Companies will need to extend identity protection and access controls to non-human users, enforce least-privilege access, monitor agent activity and safeguard the credentials that enable operation.
An AI agent with broad permissions could become an attractive target if its privileges are not tightly managed.
Cybercriminals are also using AI to accelerate the speed, scale and credibility of attacks.
The Communications Authority of Kenya recorded 11.1 billion cyber-threat incidents in the year to June 2026, a 29 per cent rise from 8.6 billion the previous year.
Distributed denial-of-service attacks jumped 114.3 per cent to 72.16 million, up from 33.68 million, while web-application attacks rose 99 per cent to 51.51 million from 25.89 million.
Attackers are employing AI to craft more convincing phishing messages, translate scams into Swahili and produce deep-fake voices and videos that mimic real people.
Such techniques can deceive employees into revealing information, authorising payments or following instructions that appear to come from colleagues or senior executives.
Ng’etich said tasks that previously took weeks now take hours because AI helps attackers find vulnerabilities faster, and that the volume of attacks on websites and other systems has increased.
Security operations centres are also adopting AI to analyse and correlate far more signals than human analysts could manually, allowing them to spot patterns across login events, file accesses and other activities.
Ng’etich explained that AI enables security teams to examine hundreds of data points, connect seemingly minor events and generate summaries and incident reports for technical or executive audiences.
He added that SOCs can use AI for vulnerability searches, enabling faster remediation of discovered flaws.
Security practitioners are deploying AI agents to handle repetitive tasks such as case summarisation, event correlation, prioritisation and standardised responses like blocking malicious elements.
The goal, according to Ng’etich, is to assist human analysts in responding more quickly and focusing on decisions that require judgement, rather than to replace them.
