What happened
Kaspersky announced (Oct 2-5) the regional launch of Kaspersky AI Protect (KAIP) across Asia Pacific: a service that inspects LLM weight files, AI-agent components and skill scripts for poisoning/malicious functionality before production use, combined with legacy threat detection, deployable via HTTP/ICAP in front of web portals, NAS, mail gateways, and as Docker/K8s/AWS EKS containers on Linux.
Why it matters
KAIP addresses the AI supply-chain gap — poisoned model weights, malicious agents and skill scripts smuggled through externally sourced artefacts — which traditional endpoint/perimeter tools miss; Kaspersky is a top-tier consumer/enterprise vendor, and APAC is its strongest region with the highest reported attack volumes (92k AI-branded attacks in H1 2026).
Applicability
APAC enterprises operating on-prem LLM/RAG/agent stacks should test KAIP coverage against their model formats and agent frameworks before adopting; format/detection-benchmark specifics are still undisclosed, so validation is required.