Think about the last time someone on your team started using an AI note-taker in meetings. A quick trial, a few sign-ups, and within a month it was quietly part of how the work gets done. Now picture that story repeating with AI assistants and agents, only faster, and often before anyone in security has seen the paperwork.
That gap is the problem. AI adoption in Kenyan organisations is moving faster than the governance meant to keep it safe, which leaves businesses exposed to old, unpatched vulnerabilities and a brand new AI-driven attack surface at the same time. Every agent, tool and assistant a company adopts is another doorway someone could try to walk through. Gartner has named agentic oversight as the most important cybersecurity trend of 2026, and it is easy to see why.
The visible edge of this risk is already familiar: investment scams built on deepfake videos, phishing and impersonation. The Nomani investment scam has grown by around 62% year-on-year globally, with more than 64,000 malicious URLs blocked in 2025 alone.
The quieter threat sits behind the scenes. Unlike a chatbot, an autonomous AI agent fetches data, follows links, downloads components and completes tasks across connected systems. It does this at remarkable speed, usually with standing permissions. If it is manipulated, misconfigured or left to run without supervision, it becomes a distinct attack surface of its own. ESET’s research has already uncovered AI-enabled threats including PromptLock and PromptSpy, and has detected thousands of malicious agentic skills circulating in the wild. An agent that can reach across the AI supply chain can carry anything on its back.
So what does a sensible response look like? First, a dedicated AI security layer built for autonomous agents. It scans AI-related files and components, inspects the external URLs agents are given, and follows the full download chain so a staged or multi-stage attack is caught before it completes. The goal is to let agents keep working without opening new doors. That is the difference between productivity gained and risk imported alongside it.
Second, AI behavioural monitoring. This watches what agents actually do, tracks their behaviour continuously and flags anything suspicious as an incident for investigation. If an agent steps outside its intended limits, early detection can prevent serious damage. Picture an agent given authority inside a bank that starts making decisions it was never meant to make.
For Kenyan businesses, an agent acting outside its guardrails is more than a technical failure. It is a compliance and accountability problem. Under the Data Protection Act, a company that suffers a personal data breach must notify the Office of the Data Protection Commissioner within 72 hours of becoming aware of it. Critical infrastructure operators face a tighter window under Section 65 of the Computer Misuse and Cybercrimes Act. The penalties are heavy, both financially and in reputation.
The takeaway is to stop treating AI as something security has never seen before, and to invest in defensive and preventative technology that understands the risks and the speed at which they move through a business. ESET has developed a platform to match Kenya’s growing appetite for AI, so companies can embrace innovation without losing ground, within an environment that keeps trust intact. Its experience across machine learning and cybersecurity means businesses get these capabilities under one roof, along with security they can rely on to assess and protect all their systems.
Never miss a story 💜
Breaking news, real stories and money tips for Kenyan women, as they happen.