Rogue Security puts agent actions on a millisecond clock

Rogue Security, a nine-person seed-stage startup, is one of 21 companies in ScaleUp House’s inaugural cohort and says its runtime platform catches malicious AI-agent actions in milliseconds. Calcalist’s cohort profile says the platform starts with coding agents and runs in-VPC with no data egress; Rogue’s product page advertises sub-5ms runtime guardrails. Builders giving agents production permissions should measure prevention latency and blocked-action coverage before treating human approval as a control, alongside the archive’s consent-budget analysis for agent workflows.