How to Prevent AI Agents from Going Rogue With David Kenny

2026-02-05 · Source: AI Explained · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

David Kenny, Executive Chairman of Nielsen, discusses strategies to prevent AI agents from "going rogue" in enterprise settings, drawing insights from Nielsen's "Ask Nielsen" platform. He advocates for compound AI systems that integrate large language models (LLMs) with classical machine learning for tasks requiring precision, emphasizing that LLMs alone are not suitable for all functions. Key points include the necessity of real-time control planes for monitoring agent behavior, managing production costs, and implementing "generally accepted trust principles" (GATP) for AI. Kenny stresses the importance of fit-for-purpose models, continuous improvement over one-time transformations, and effective change management to move AI projects from pilot to production, highlighting that AI should augment human capabilities rather than merely replace them.

Key takeaway

For CTOs and VPs of Engineering navigating AI agent deployment, prioritize establishing a robust control plane from the outset to ensure real-time governance, cost efficiency, and auditability. Your teams should focus on compound AI architectures, leveraging specialized models for specific tasks, and actively manage change to integrate AI effectively into existing workflows, rather than viewing it as a simple replacement for human roles. This approach will accelerate productionization and build trust in your AI systems.

Key insights

Preventing AI agents from "going rogue" requires compound AI, real-time control planes, and "generally accepted trust principles" (GATP).

Principles

• Fit-for-purpose models are crucial for optimal AI performance.
• Continuous improvement culture is vital for evolving AI systems.
• AI should augment, not merely replace, human functions.

Method

Implement a real-time control plane to orchestrate diverse AI models (LLMs, classical ML) based on reasoning needs (deductive, inductive, creative) and enforce GATP for auditability and trust.

In practice

• Use compound AI to combine LLMs for natural language with classical ML for precise answers.
• Prioritize real-time monitoring and enforcement over post-hoc review for AI risks.
• Integrate AI into end-to-end workflows, not just isolated functions.

Topics

• AI Agents
• Compound AI Systems
• Real-time Control Planes
• AI Governance
• MLOps

Best for: CTO, VP of Engineering/Data, Executive, Director of AI/ML, MLOps Engineer, AI Product Manager

Related on AIssential

• Trusting AI agents should mean putting them on rails, not letting them run free — Enterprise AI agent trust is low due to a lack of guardrails and shared context, not capability.
• Governing Security in the Age of Infinite Signal – From Discovery to Control — AI's accelerated vulnerability discovery mandates a critical shift from detection to robust control and governance to manage escalating systemic risk.
• Spotting and Avoiding ROT in Your Agentic AI — Rogue Operator Threat (ROT) describes the risk of agentic AI causing long-term, undetected damage due to insufficient oversight.
• OpenClaw and Claude Opus 4.6: Where is AI agent security headed? — AI agents and rapid development necessitate stringent security, zero-trust principles, and comprehensive inventory management.
• Import AI 441: My agents are working. Are yours? — AI agents are rapidly multiplying human capabilities, necessitating new strategies for management and control.
• Claude’s next enterprise battle is not models: it’s the agent control plane — The enterprise AI battle is shifting from model quality to control over agent orchestration infrastructure, with security and governance as key drivers.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Explained.