Definition
A network security technique that isolates individual AI and machine-learning workloads — such as a specific LLM inference container — from each other at the network level, so that a security breach in one AI service cannot automatically spread to others sharing the same infrastructure. Traditional network rules treated all containers in a cluster as equivalent; microsegmentation applies fine-grained rules based on what the container actually does and which AI model it is running. This closes a gap where attackers who compromise one AI model could pivot to adjacent models or data stores.
Why it matters
As organisations run multiple AI models and agents on shared cloud infrastructure, microsegmentation prevents a single compromised AI workload from becoming a launchpad for broader infrastructure compromise — a critical containment control for enterprise AI deployments.