RESEARCH · THREAT INTELLIGENCE

Field notes from the adversary's side

The place practitioners come for real vulnerability research — how attacks actually work, what we're seeing across AI and agent systems, and what it means for the standards you report on. Written by the operators who run the engagements.

Read the latest Subscribe

FEATURED

  • White paper
  • AI & LLM

Indirect prompt injection: the attack path traditional pen tests never see

How a single poisoned document turns a helpful assistant into an exfiltration channel — and the trust-boundary controls that close the path. Mapped to OWASP LLM Top 10 and MITRE ATLAS.

Format
White paper · 18 pages
Topic
AI offensive security
Frameworks
OWASP LLM · ATLAS
Read the paper

LATEST RESEARCH

What we're publishing

Vulnerability write-ups, threat intel, and analysis from live engagements.

  • Blog

Chaining two tools into an admin action

An agent tool-abuse walk-through: how low-privilege input becomes privileged execution — without touching the admin UI.

Read
  • Threat intel

FedRAMP red team: what the draft guidance changes

Red team exercises as an organizational control under NIST 800-53 Rev. 5 — what to expect, and how to prepare now.

Read
  • Blog

RAG pipelines are an attack surface

Vector-DB poisoning and embedding inversion in practice — and why retrieval security belongs in your threat model.

Read
  • White paper

Proof over theory: what a real exploit chain looks like

An anonymized engagement teardown — recon to reproduction — showing business impact, not just CVSS.

Read
  • Podcast

Override, episode 01: becoming the adversary

How our operators think about pre-emption — running the attack before there's an attacker on it.

Listen
  • Video

Live jailbreak: multi-turn escalation on a copilot

A recorded demonstration of incremental prompt escalation, and the guardrail design that stops it.

Watch

WHERE TO FIND IT

Four ways we share the work

Whichever way you learn best, the research is the same — written and recorded by the operators who run the engagements.

  • AI & LLM security
  • Prompt injection
  • Agent & tool abuse
  • Red team
  • FedRAMP
  • Cloud
  • Application security
  • Threat intel
// 01

Blog

Short, technical write-ups on vulnerabilities, attack paths, and what we're seeing in the field.

// 02

White papers

Deeper analysis and methodology — the reference material your team can cite internally.

// 03

Video

Recorded demonstrations of real techniques, from prompt injection to lateral movement.

// 04

Podcast

Conversations with operators on how modern attacks — and modern defenses — actually work.

Get the research before the adversary does

New vulnerability write-ups, threat intel, and engagement teardowns — straight to your inbox.

Subscribe to Override research