“Rubrik's core thesis is that securing and governing AI agents requires using AI, specifically small language models, to meet latency and cost requirements.”