AI-Driven Data Centres Demand Smarter Security and Faster Deployment
Alexandre Depret-Bixio, SVP International at Anomali, discusses why trusted data is becoming the foundation of secure AI operations, how organisations are breaking down silos between infrastructure and security, and why vendors are embracing augmentation-first deployment models to help enterprises modernise AI environments more quickly and with lower risk.
What new security risks emerge as data centers become more software-defined and AI-orchestrated?
The risk that gets underestimated alot these days is speed. When infrastructure was human-operated, a bad decision took hours to cause real damage, which gave someone time to notice and step in. When AI agents are making decisions inside a software-defined environment, a bad decision can spread across the entire system in seconds.
The root cause is usually the same. Agents end up acting on data they don’t actually understand. Give an AI system raw, unvalidated signals and it will make mistakes quickly and at scale. That’s the real new risk emerging right now.
The fix is making sure every agent has validated, trustworthy context before it acts, and that every action it takes can be explained, reversed, and traced back afterward. Autonomy without that kind of governance just means risk moving faster.
Where do you see the convergence between physical infrastructure management and cybersecurity today?
We’d frame this as a data convergence problem more than a physical one. As infrastructure gets more distributed and dynamic, security teams are dealing with telemetry from more places than ever, spanning cloud, on-prem, IoT, and OT, and much of it never gets reconciled into one picture. The systems that manage infrastructure and the systems that secure it have historically spoken different languages.
That’s starting to change. Organizations ahead of the curve are pulling that signal into a single, unified layer that’s normalized, deduplicated, and enriched with context, so it can be searched and acted on in seconds. Once infrastructure and security data live in the same place, security becomes part of how the infrastructure runs day to day, rather than something added on afterward.
How are vendors adapting to the need for faster deployment cycles in AI infrastructure environments?
The old model involved convincing a customer to tear out what they have and start over, which tends to be slow, risky, and expensive. Few organizations have the patience or budget for a multi-year overhaul while threats keep moving faster.
The shift we’re seeing is toward augmentation-first deployment. Vendors are designing products to sit on top of existing systems, prove value within weeks, and let customers expand at their own pace. It’s also a more honest model, since it asks vendors to earn the right to go further by proving value early.
