Kernel-level ransomware prevention and AI usage governance for the endpoint

Praegis is an early-stage company launched in 2025 and based in Malibu, California, building on technology it markets as Mahiri. Its premise is that the two fastest-moving risks on a corporate endpoint now sit next to each other: ransomware that executes faster than any recovery process, and employees pasting regulated data into whatever model is convenient. The AI governance side inventories AI tool use, blocks the tools an organization has not sanctioned, enforces guardrails on the models it has approved, and monitors and encrypts sensitive content continuously rather than sampling it at a gateway.

The ransomware side operates at the kernel and is behavioral rather than signature-driven, intercepting processes at what the company calls the moment of intent, before files are encrypted, data is moved, or operations are disrupted. Praegis states that the behavioral intelligence underlying this detection derives from NSA research; that attribution is the company's own. Both products share the same posture, which is preemptive enforcement on the device rather than detection followed by response, and both are aimed at organizations that have adopted AI tooling faster than they have built policy around it. Praegis has not published funding details or a named leadership roster.