AI Explained How to Prevent AI Agents from Going Rogue With David Kenny
Feb 6, 2026
David Kenny, Executive Chairman of Nielsen and veteran AI leader, discusses preventing AI agents from going rogue. He covers control planes and real‑time enforcement. He talks about fit‑for‑purpose and compound AI, trust principles and audit readiness. He explains balancing safety controls with developer velocity and moving pilots into production.
AI Snips
Chapters
Transcript
Episode notes
Design For Continuous Velocity
- Do build a culture of constant improvement that works in short cycles and accepts self-learning systems.
- Do adapt organizational velocity rather than one-off transformations to keep pace with creative AI advances.
Control Plane Is Nonnegotiable
- Observability alone isn't enough; you need a control plane with ethics, rules, and runtime checks to prevent rogue agents.
- Without that control plane you risk uncontrolled, unexpected, and harmful results from agents.
Match Models To Tasks
- Do use the right model for each task: LLMs for language, classical ML for precise known answers, and other models for deduction or creativity.
- Do orchestrate models via a control plane so each component uses the fit-for-purpose model.