Artificial Intelligence (AI) has moved far beyond experimentation and is rapidly becoming part of everyday business operations. Employees are using AI assistants to draft content, analyse information, automate repetitive tasks and improve productivity. At the same time, developers and business teams are integrating AI-powered tools, plugins, agents and workflows into existing systems.
While these technologies can deliver significant efficiency gains, recent research suggests they may also introduce new cybersecurity risks that many organisations have not yet fully considered.
What the Research Found
According to ESET, researchers analysed nearly 900,000 AI skills from public repositories during the first half of 2026. Their analysis identified approximately 25,000 suspicious skills and more than 3,000 malicious examples.
ESET reported that some of these malicious skills incorporated tools associated with credential theft, remote administration and persistence techniques. Researchers also identified AI-related security tools that appeared legitimate but failed to provide meaningful protection, potentially creating a false sense of security.
These findings highlight the emergence of a new challenge: the AI ecosystem itself is becoming part of the enterprise attack surface.
AI Risks Extend Beyond Chatbots
Many organisations have focused their AI security discussions on concerns such as employees entering confidential information into public AI services or relying on inaccurate AI-generated responses.
While those concerns remain important, modern AI environments are becoming increasingly complex and may include:
• AI assistants and chatbots
• AI-enabled business applications
• Third-party AI plugins and integrations
• AI skills and extensions
• Autonomous or semi-autonomous AI agents
• AI coding assistants
• Custom workflows connected to business data
Each new integration may introduce additional permissions, dependencies and potential vulnerabilities that organisations need to understand and manage.
Shadow AI: The Visibility Problem
One of the key themes in ESET’s article is that employees themselves are often not the primary risk. Most staff adopt AI tools because they are attempting to work more efficiently and solve business problems faster.
The greater concern is often a lack of visibility.
An employee may connect an AI service to a business application, install an AI extension or adopt an external AI platform without the knowledge of IT or security teams. This practice, commonly referred to as “Shadow AI”, can create security, compliance and governance challenges because organisations may not understand what information is being accessed, stored or processed.
Why This Matters for South African SMEs
For many South African small and medium businesses, AI presents an attractive opportunity to improve productivity without significantly increasing headcount. However, SMEs frequently operate with limited IT resources and often lack dedicated governance frameworks for emerging technologies.
Without proper controls, AI tools may gain access to:
• Customer information
• Financial records
• Employee data
• Contracts and legal documents
• Business intellectual property
• Cloud-based collaboration platforms
This becomes particularly important in environments where Microsoft 365, SharePoint, Teams, OneDrive and line-of-business applications are interconnected.
IT Evalution Analysis
Source Fact: ESET’s research demonstrates that malicious and suspicious AI skills exist within public AI ecosystems and that AI integrations can expand an organisation’s attack surface.
IT Evalution Analysis: The primary risk for most South African SMEs is unlikely to be advanced AI malware itself. The more immediate concern is unmanaged adoption of AI services without governance, visibility or security review.
As businesses increasingly connect AI tools to Microsoft 365 environments, CRM platforms, accounting applications and business databases, organisations should treat AI services with the same level of scrutiny applied to any other cloud application or third-party integration.
Governance Is Becoming Essential
The question for most businesses is no longer whether employees should use AI.
Instead, organisations should ask:
• Which AI tools are being used?
• What business information can they access?
• What permissions have been granted?
• Where is data processed and stored?
• Are usage policies defined and understood?
Businesses that answer these questions early are likely to gain the productivity benefits of AI while reducing operational, security and compliance risks.
Contact Us
“This article has been prepared by IT Evalution using publicly available information from the sources listed below. The recommendations and business considerations represent IT Evalution’s independent analysis and interpretation.”
Primary Source
Organisation: ESET
Article Title: Are Employees Creating Your Organisation’s Next AI Security Risk?
Author: Not specified in the published article
Publication Date: 2026 (publication date not explicitly stated in the article page reviewed)
Source: View Original ESET Article
