Insights
Essays, analysis, and careful critique on what works, what fails, and why.
What
We publish analysis on the mechanics of security failure and the reality of adversary tradecraft. This is not marketing fluff or vendor-speak. It is operational observation.
So What
The security industry is drowning in noise. We filter for signal. If it doesn't help you stop an attack or explain why a control failed, we don't write about it.
Now What
Read these to challenge your assumptions. Use them to ask better questions of your vendors, your teams, and your own architecture.
Agents Lack Judgment: The Rules Problem in Agentic Penetration Testing
Kevin Mandia said it on CNBC: AI agents lack judgment, so they need definitive rules. He is right about the diagnosis. Rules are where the argument starts, because somebody with judgment has to write them.
Indirect Prompt Injection Testing: Test the Data Paths, Not the Prompts
In production, the hostile prompt never arrives as a prompt. It arrives as a tool result, a retrieved document, an email the user asked the agent to summarize. Indirect prompt injection testing means testing the data paths: who can write to each channel, what the agent does with a planted payload, and how far the chain gets. The corpus is the unit test. The engagement is the integration test.
What AI Pentest Agents Get Wrong
The autonomous pentest agents are genuinely impressive. They are also wrong about what the job is. Speed and coverage are real advantages. They are not the same thing as adversary reasoning, and the vendors selling replacement are measuring the wrong thing.
MCP Tool Poisoning: Turning the Agent's Tools Against It
Tool poisoning is the attack class MCP was born vulnerable to: a tool description the agent trusts, carrying instructions the operator never wrote. Here is how the class works, what it looks like in the wild, and how to test for it.
AI-Enabled Pentesting vs AI Pentesting: The Terminology Split
AI pentesting is two different engagements wearing one name. In one, the AI is the tool. In the other, the AI is the target. Buyers are comparing quotes for different products.
How to Penetration Test an AI Agent
Most AI pentests test the chatbot and call it the agent. An agent is a system: model, tools, memory, orchestration, identity. Here is how to test the whole thing.
What Is Agentic Pentesting?
Agentic pentesting means AI agents that plan, use tools, and adapt during a test. Most of the demos I've seen are automation wearing a costume. Here is how to tell the difference.
The Agentic AI Security Testing Gap
LLM scanning has 176 probes. MCP has 16. A2A: zero you can own. A CISO breaks down the agentic AI security tooling gap and what he is building to close it.
Tradecraft is Decision Logic
Tradecraft isn’t a toolset or a MITRE checklist. It’s the decision logic attackers use to navigate your environment.
The Validation Gap
Why point-in-time security assessments create false confidence - and how continuous, methodology-driven validation closes the gap.
Compliance is Not Security
Compliance proves you meet a baseline. Security proves you can survive an adversary. Confusing the two creates expensive, avoidable risk.
Content Pillars
We focus our writing on four distinct themes. These are not random; they are the fault lines where we see security programs continuously fail.
Validation Failure
Why passing an audit is not the same as being secure, and why "green" dashboards often hide critical weaknesses.
Why this matters: Most organizations optimize for the appearance of security (compliance) rather than the reality of resistance. Decision makers need to know why their investments aren't stopping breaches.
Adversary Tradecraft
Observed attacker behavior, methodology analysis, and how determination defeats static playbooks.
Why this matters: You cannot defeat what you do not understand. If your validation model is based on a static list of CVEs, you will lose to an adversary who operates on decision logic and OODA loops.
Constrained Environment Operations
How to hunt and validate in environments where visibility is poor, telemetry is fixed, and tools cannot be easily changed.
Why this matters: Not everyone has a blank check or a perfect stack. Leaders need strategies to win with the army they have, not the army they wish they had.
Security Theater
The courage to stop doing things that look like work but produce no evidence of risk reduction.
Why this matters: Resources are finite. Every dollar and hour spent on theater is one stolen from actual defense. We highlight these waste centers to help you reclaim your budget.