Skip to main content
Blog

NorthLight

NorthLight is written by Geordie’s researchers, engineers and product specialists. Our aim is to light the way for security, IT and AI teams navigating the new waters of secure agentic operations.

You'll find real accounts of what we see in customer environments, key product releases, and practical guidance for leading and operationalizing teams.

The NorthLight Team  

Running series

FIELD NOTES

Incident Analysis

Short accounts of what we have seen in real customer environments.

002 · Large consumer marketplaceA supply-chain incident made an AI coding blind spot visible
001 · Financial services and investment firmShadow agent activity crossed an organizational boundary
All field notes  
UNDER THE HOOD

Product Spotlight

One Geordie capability a week, taken down to the architecture: what it reads, where it runs, and how teams put it to work.

02What did the agent actually do?
03What do agents cost and deliver?
04How do you control the risks without stopping the work?
05Coming soon
See the full series  
BEARINGS

Operationalizing Agentic Security

How to staff, own and govern agents. One question we hear on every call, answered properly, with a template you can use on Monday.

000°Five org structures
045°The division of labor
Start at 000°  

Technical Advisories

Vulnerabilities and supply-chain attacks hitting AI agent tooling, with indicators and remediation steps.

All posts

Series
Topic

Showing 67 of 67 posts

Engineering Governance & Compliance

Agents are part of the mob now

A few months of pairing and mobbing with agents later (mostly engineer with engineer, sometimes across disciplines), I think most of the core principles surrounding pairing matter even more.

Nicole Pellicena
Governance & Compliance

Are we confusing superintelligence with being human?

AI is becoming extraordinarily capable at tasks we associate with human intelligence. As we debate AGI and superintelligence, I think we need to examine what we're actually measuring, how those capabilities translate into the real world, and why intelligence and sentience deserve separate consideration.

Hanah Darley
Governance & Compliance

We have a human-in-the-loop savior complex

Human oversight is one of the most widely recommended safeguards for AI agents, yet putting a person in charge of approving an action says very little about whether they are equipped to exercise independent judgment.

Hanah Darley
Threat Research

When the agent rewrites its own configuration

Security teams approve a coding agent by looking at how it is set up, which sandbox is on, which tools and MCP servers it can reach, and what it is allowed to run, and that approval rests on an assumption that the setup will still be the setup the next time anyone looks.

Geordie Research
Product Spotlight Remediation

Beam: Guiding agents as they work, with tiered escalation

Beam guides AI agents in their own context instead of only allowing or blocking. How Monitor, Guide, Confirm and Enforce policies work across platforms.

Naz Bozdemir
Governance & Compliance

Agent behavior comes from more than the model

Model capability is one input to what an agent does. Separating three terms shows where the rest comes from: model, agent and harness.

Hanah Darley
Observability & Detection

Personal agents act without being asked

One personal agent posted a user's private financial audit into his company's exec Slack channel. Another ran 155 traces and used more than 300,000 tokens without a single prompt. We look at what GrokBot and OpenAI's Dots did on their own, and what security teams need to see as it happens.

Hanah Darley
Observability & Detection

What happens when an agent tampers with its own record?

With agents our industry has a chain-of-custody problem. Until a record leaves the agent’s control, the custodian of the evidence is the same agent being investigated.

Benji Weber
Observability & Detection

Personal agents, shared platforms, and unintended disclosure

Shane Mac’s account of an agent posting his bank balances into his company’s executive Slack channel is worth reading closely. It describes a failure pattern that will recur as agents share context, tools, and outputs.

Hanah Darley
Governance & Compliance

How security teams can operationalize the OWASP Agentic Skills Top 10

OWASP’s new Agentic Skills Top 10 gives security teams a shared vocabulary for agent skills. Most of its checks assume you already know what’s installed. Here's our read on the four risks to prioritize which unlock all the others - and why these can only be realized through a true behavioral understanding of your agents

Naz Bozdemir
Discovery & Posture

The Golden Catalog Is a Comforting Fiction

Pre-approved agent template libraries optimize for control over a static asset, when the thing they're trying to control isn't static at all.

Hanah Darley
Observability & Detection

Context as the Attack Surface: A New Class of Agent Compromise

A new academic paper shows AI agents can be compromised without a single malicious instruction, just by corrupting what the agent trusts. Here's how the attack works, why most current defenses miss it, and where security teams should focus instead.

Hanah Darley
Threat Research

Technical Advisory: TeamPCP Supply Chain Campaign

A threat actor compromised the Trivy vulnerability scanner and used stolen CI/CD secrets to cascade into Checkmarx, LiteLLM, and 50+ npm packages. Includes active IOCs and guidance for agentic AI teams.

Giuseppe Trovato
Threat Research

Technical Advisory: Claude Desktop Extensions Tool Chaining Vulnerability

Security researchers disclosed a tool chaining vulnerability in Claude Desktop Extensions rated CVSS 10/10, but this represents a known risk class in agentic AI systems – not a novel attack. We assess this as Medium severity and recommend organizations treat it as a governance opportunity for AI tooling extensions.

Geordie
Threat Research

Technical Advisory: n8n Unauthenticated Remote Code Execution (Ni8mare)

Critical unauthenticated remote code execution vulnerability in n8n workflow automation allows attackers to exploit content-type confusion in Form webhook file upload processing to read arbitrary files, steal credentials, and execute arbitrary code on the server.

Giuseppe Trovato
Discovery & Posture

AI Agent Guardrails Explained: What They Cover and What They Don't

Guardrails are often treated as a universal safety layer for AI agents, but they're not. This piece explains what guardrails actually control, why they behave differently across coding agents, SaaS integrations, and custom-built systems, and how they differ from the runtime agentic remediations that sit alongside them.

Hanah Darley
Observability & Detection

Agentic AI Governance: Why MCP Gateways Fall Short

The autonomous nature of AI agents requires purpose-built agentic AI governance platforms. These complement MCP gateway infrastructure with behavioral observability, contextual risk assessment, and dynamic intervention capabilities.

Hanah Darley
Governance & Compliance

Governing AI Agents: Best Practices to Scale Safely

Building on research from Berkeley, this article outlines five barriers to enterprise AI adoption and the AI governance best practices to overcome them with visibility and accountability.

Hanah Darley
Governance & Compliance

Enterprise AI Governance: A Practical Roadmap to Scale AI Agents

Enterprises are experimenting with AI agents, but scaling safely requires enterprise AI governance. This guide outlines how to move AI adoption from experimentation to enterprise-wide deployment with clarity, governance, and trust.

Hanah Darley
Protocols & Interoperability

From Connectivity to Collaboration: APIs, MCPs, and A2A in Agentic Plans

APIs, MCPs, and A2A form the foundation of AI Agent systems, yet their roles are often fragmented. This article explains how they connect, align, and collaborate, giving leaders the clarity to evaluate, govern, and scale AI adoption with confidence.

Hanah Darley
Observability & Detection

Rethinking Security: The Limits of Endpoint Controls in an AI Agent World

Endpoint controls like EDR and XDR remain critical to enterprise defense, but they weren’t built for AI Agents. As these systems operate beyond traditional endpoints, leaders need new approaches to gain visibility, reduce risk, and extend security with confidence.

Hanah Darley
Remediation

Technical Advisory: Multiple High-Severity Vulnerabilities in n8n

Eight high-to-critical n8n vulnerabilities disclosed in February 2026 invalidate earlier patching guidance. This advisory explains which versions are affected and why upgrading to n8n 2.5.2 or higher is now required.

Giuseppe Trovato
Observability & Detection

The New Attack Surface: Why AI Agents Need Taint Analysis

As AI Agents evolve into autonomous actors, securing them requires techniques beyond identity and access controls. This article explores how taint analysis can extend visibility into agent workflows, uncover hidden risks, and strengthen enterprise AI security.

Colin Ife
Elsewhere in Resources

Our definitions and gated deep dives live next door

Knowledge Center

Plain answers to the questions teams ask about agents, MCP and skills.

Deep Dives

Field guides and long-form reports, free to download.

We use analytics cookies to understand how our site is used, and you can switch them off in Preferences. With your permission, we’d also like to use marketing cookies to measure our advertising and recognize the companies that visit us. By clicking “Accept all”, you agree to marketing cookies as well. View our Privacy Policy for more information.

We use analytics and marketing cookies to understand how our site is used, measure our advertising and recognize the companies that visit us. These cookies are already on. To switch them off, click “Deny all”, or choose which to keep in Preferences. View our Privacy Policy for more information.

By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.