— RAE RESEARCH & WRITING

Adversarial intelligence, documented.

Attack research, detection methodology, and field notes from defending enterprise AI agents.

— LATEST

2026 — COMING SOON

STORMRAE: What 64,634 adversarial prompts taught us about prompt injection

A post-event analysis of the largest coordinated red-team exercise ever run against a deployed enterprise AI system. What we saw, what held, and what didn't.

Read →

2026 — COMING SOON

The five detector heads: how RAE reaches consensus

RAE requires 3 of 5 independent detectors to agree before any blocking action fires. Here's why that number matters, and how we calibrated each head.

Read →

2026 — COMING SOON

Why raw prompts must never leave your network

The architecture case for on-premises prompt handling — and why metadata-only cloud telemetry is the right split for enterprise deployments.

Read →