AI is becoming both a weapon and a shield. At Data Centres in the AI Era, Rob Hankin, CTO at Cybit, explains how operators can safeguard critical infrastructure when large-language models make advanced exploits cheap and scalable:

• AI-powered attack automation – polymorphic malware, self-tuning DDoS campaigns and code-generation tools that let hobbyists mount enterprise-grade assaults.
• Reconnaissance at machine speed – LLM-driven vulnerability discovery that crawls public-facing portals, legacy APIs and customer workloads to build real-time exploit maps.
• Operational-technology weak points – why ageing PLCs, HVAC and power-controls exposed for remote maintenance are now prime lateral-movement targets.
• Zero-trust remote access & segmentation – hardening out-of-band interfaces, enforcing least privilege, and containing breaches before cooling, power or door-control cascades occur.
• AI-enabled cyber-defence – deploying machine-learning analytics, red-team simulations and predictive threat modelling to fight AI with AI.
• Regulatory readiness & customer education – aligning with the forthcoming EU AI Act, documenting AI security practices, and training tenants to avoid inadvertent data leakage into public models.