Akamai Research Warns Top 5% of Enterprise AI Users Pose Disproportionate Security Risk

According to research published by Akamai, a small cohort of enterprise AI power users—comprising the top 5% of users—are posing a disproportionate security risk by hardcoding unvetted AI tools into critical business operations. While enterprise security teams have been focused on policing the proliferation of general employee usage of tools such as ChatGPT and Claude for drafting tasks, the study identifies a more urgent threat: a handful of super-adopters who are embedding unvetted AI capabilities directly into mission-critical systems.

The findings suggest that these power users are integrating AI tools beyond approved channels, effectively hardcoding them into operational workflows. This practice creates potential vectors for data leakage, compliance violations, and the introduction of security vulnerabilities into essential business processes. The research contrasts the risk profile of these elite users with the broader base of employees using AI for auxiliary tasks, emphasizing that the concentration of risk among a small user segment carries outsized organizational impact.

Akamai researchers note that the risk stems from policy gaps and shadow IT patterns rather than a specific active exploit in the wild. The integration of unvetted AI capabilities into critical infrastructure raises concerns about data exfiltration, unauthorized access, and compliance failures, particularly in regulated industries. The report does not detail specific active exploitation but highlights the risk as a consequence of inadequate governance over advanced AI adoption.