OpenAEV v3: Adversarial Exposure Validation Goes Autonomous with Attack Chaining

Security teams have spent the last decade getting good at knowing what could hurt them. Threat intelligence programs have matured, MITRE ATT&CK is now a shared vocabulary, and most organizations can produce a credible list of techniques an adversary targeting their industry is likely to use. But all that intelligence still hasn't answered the question CISOs face in every boardroom today: can you prove we're secure? That's the question adversarial exposure validation is built to answer.

TL;DR

• Attack Chaining introduces a new type of scenario that turns OpenAEV into a live, adaptive attack path engine, where the output of one action feeds the input of the next.
• A new Exposure Command Center and Adversarial Exposure Score give SOC leaders a single-glance, quantified, and actionable view of posture and resilience to prioritize actions.
• AI Red-Teaming extends validation to LLMs and AI agents, testing for prompt injection, jailbreaks, and tool abuse alongside existing EDR, SIEM, and email defenses.
• OpenAEV v3 also adds automated reporting, asset criticality scoring, new integrations, expanded social engineering coverage, and multi-tenancy enhancements, part of a broader set of platform updates.

This gap between knowing and proving is only widening: Filigran’s 2026 State of Threat Management report finds that 93% of security leaders struggle to maintain an accurate view of their attack surface, 84% have been hit by attacks exploiting risks they already knew about, and 88% agree point-in-time assessments can't keep pace with how fast systems and environments now change.

This is the gap OpenAEV was built to close. While OpenCTI democratizes threat intelligence as the world's most widely adopted open-source CTI platform, OpenAEV extends that mission into validation: giving security teams a way to test whether their defenses hold up in practice.

Since launch, it's grown from a community-driven breach and attack simulation tool into an enterprise-scale, agentic Adversarial Exposure Validation platform, staying rooted in the same open-source, threat-driven DNA as the rest of Filigran's eXtended Threat Management (XTM) platform.

With v3, OpenAEV takes its biggest step forward yet.

Why OpenAEV Is Changing

Everything shipping in v3 is pointed at the same underlying shift: Security validation needs to stop being a periodic, manual exercise and start being something continuous, automated, and increasingly autonomous.

Concretely, that means OpenAEV is becoming:

• More actionable, giving teams a clear, prioritized view of what to fix rather than a long list of disconnected findings,
• More automated, replacing manual scripting and one-off engagements with simulations that can be run on demand and repeated as often as needed, and
• More agentic, handing an increasing share of the planning and execution of a validation exercise to AI agents that can adapt to what they find in real time, the same way a skilled red teamer would.

Those three shifts show up everywhere in this release. They're why OpenAEV now has a dedicated Exposure Command Center instead of a scattered set of dashboards, why there's now a single Adversarial Exposure Score to track resilience over time, and why the platform can, for the first time, run genuinely autonomous pentesting from a plain-language objective through to a validated result. All of it is built for the same purpose: helping organizations match a threat landscape that is moving faster, and increasingly thinking for itself, without asking security teams to scale headcount to keep up.

And the centerpiece of that effort, and the feature this release is really built around, is Attack Chaining.

Attack Chaining: Testing Like an Attacker, Continuously

Real adversaries, increasingly AI-powered ones, don't run multiple isolated, disconnected actions against a target. They chain them: recon reveals a target, a credential dump reveals a password, that password unlocks the next machine, and so on, with each step depending on evidence the last one uncovered.

Scripted simulations and atomic tests still surface exposures, but they can't replicate adaptive human or AI behavior. A pentest or red-team exercise can recreate it manually, but that's costly and, more importantly, slow and periodic. What's more, modern pentests focus on technical testing, forgetting about the leading cause of breaches: people.

Attack Chaining is OpenAEV's answer to that problem. It introduces a new type of scenario that turns the platform into a live, adaptive multi-stage attack path engine, where the real output of one action (a credential, an open port, a token, a misconfigured permission) is captured as a structured finding and fed automatically into deciding what gets attacked next.

The result: an attack that unfolds and branches in real time, on an interactive graph, based on what it finds in your environment.

Beyond the obvious value of seeing and interacting with a real attack path an adversary could take, Attack Chaining gives teams a fast, realistic, and cost-effective way to test vulnerabilities at scale and pinpoint the chokepoint to prioritize, breaking an entire attack path with a single fix.

Teams have practiced attack path management for years. But OpenAEV flips the script: instead of pre-defined, vendor-locked scenarios, you're in full control to model any attack you want, orchestrated manually or run fully autonomously by agents. Here's how four capabilities work together as a single system, turning OpenAEV into an autonomous pentesting solution.

1) Open, Conditional Attack Chaining Logic

Rather than being locked into a vendor-defined scenario, teams can build reusable multi-stage attack path logic from scratch, drawing on any actions or events in OpenAEV. Conditions determine what happens next, so a chain branches automatically based on what each action discovers in the moment, and that logic can be revisited and updated at any time as new techniques emerge.

2) Live Attack Path Mapping

Teams watch each attack path unfold and branch dynamically in real time, based on findings the simulation discovers along the way, rather than reading about it only after the simulation has finished. Every hop and pivot renders live on an interactive graph from the first action to the final objective, giving a real-time view of how the path is forming rather than a summary of where it ended up.

Attack path mapping, running in real time

3) Transparent, Actionable Findings That Surface Chokepoints

Every result a chain produces, whether it's a credential, an IP address, a token, or a file, becomes a structured finding that can be drilled into directly: why the action ran, what it returned, and how that finding drove the next step. That traceability is what makes it possible to identify chokepoints (the single points in a path where one fix collapses everything downstream) instead of leaving teams with a long, undifferentiated list of things to remediate.

Actionable nodes in the attack map and chokepoint identification

4) Scope and Safety Controls

None of this runs unsupervised by default. Every chain operates inside guardrails defined upfront: which assets and targets are in bounds, which actions are permitted, and how far a chain is allowed to escalate before it stops. That's what makes it possible to run these simulations with confidence, even at scale: full autonomy in execution, without losing control over the boundaries.

Attack path logic scoping

These four capabilities are the foundation, but Attack Chaining can be put to work in two different ways, depending on how much of the process a team wants to hand off.

Operator-Led vs. Autonomous Attack Chaining

Attack Chaining can be deployed and managed manually, fully orchestrated by AI agents end-to-end, or as a combination of both with human-in-the-loop.

With operator-led Attack Chaining, a human builds the logic and controls execution, mapping out the conditional logic. Deterministic and transparent, best for scaling pentesting and validating attack paths with precision.

With Autonomous Attack Chaining, that judgment is handed to an AI agent, powered by XTM One. An operator defines an objective and a scope, and a dedicated orchestrator AI agent plans the attack path, executes it, and adapts as it goes: reacting to findings, reordering steps, selecting the right actions based on what it discovers, and even generating realistic phishing emails and landing pages when the objective calls for social engineering. The orchestrator can also call on specialist agents for payload creation, code generation, recon, and exploitation, using OpenAEV's built-in agents or bringing its own from XTM One.

Agentic orchestration of an Attack Chaining scenario, with agent scope definition

Either mode delivers the same value: a realistic, end-to-end attack simulation that runs in minutes instead of weeks, at a fraction of the cost, repeatable as often as the environment changes. And because Attack Chaining is threat-intel-native, every chain is prioritized around techniques relevant to your organization, with results feeding straight into the score below, turning a validated path into part of a unified risk measure, not an isolated report.

The Exposure Command Center: A Home Base for Prioritization

The most visible change in OpenAEV v3 is a redesigned interface, but it's more than cosmetic: it moves OpenAEV from a simulation tool into a genuine operations cockpit for exposure validation.

At its center is the Exposure Command Center, now the platform's home dashboard: a single, at-a-glance view of posture across every security domain, recent simulation results, and detection coverage mapped against MITRE ATT&CK and ATLAS.

Your Exposure Command Center for a single, actionable view of where you stand

It's designed to work for more than one audience at once: a SOC leader can use it to understand where the organization stands overall, while blue teams, red teams, and MDR practitioners can use the same view to run adversarial exposure validation and drill straight into the areas that need attention right now.

Adversarial Exposure Score: A Single Number to Manage Resilience

To ensure proper actionability and to guide an effective exposure management workflow, OpenAEV v3 introduces the Adversarial Exposure Score, visible in your Exposure Command Center, representing your exposure levels.

“You can’t manage what you can’t measure,” as the saying goes, and this is exactly what the Adversarial Exposure Score was built for: to give CISOs and practitioners a single, high-level score to measure, understand, manage, and improve their resilience over time.

The score is based around three core principles:

• It’s transparent: composed of posture scores across all main exposure domains, weighted by criticality.
• It’s actionable: You can drill down into failed and breached validations.
• It's dynamic: It will change and evolve over time, based on the results of simulations teams have run.

The Adversarial Exposure Score: a high-level breakdown

AI Red-Teaming: Testing AI Posture Alongside Human and Endpoint Posture

Most exposure validation tooling still assumes humans and endpoints are the ones making the decisions worth testing. That's increasingly wrong, as organizations are deploying AI agents faster than they can answer what happens if one gets manipulated.

OpenAEV now includes new native injectors in the Threat Arsenal under an AI Red Team category, running adversarial exposure validation against LLM models and AI agents the same way existing injectors validate EDR, SIEM, or email defenses. They test for prompt injection, jailbreaks, excessive agency and tool abuse, and Model Context Protocol (MCP) tool poisoning, using engines like NVIDIA Garak, Microsoft PyRIT, and Promptfoo. These attacks map to MITRE ATLAS, giving AI techniques the same structured coverage as any other domain. It's backed by a new Artificial Intelligence domain in the platform, so AI posture rolls into the same findings, scoring, and Command Center as every other asset.

Pre-built, out-of-the-box AI Red Team integrations, with their individual attack patterns

Automated Reporting: From Simulation Results to a Stakeholder-Ready Report

Simulation results only create value once they reach the people who act on them: compliance, leadership, the board. Getting there has traditionally meant pulling data manually across tools and reshaping it into something presentable.

OpenAEV v3 introduces a full automated reporting module that consolidates simulation, exposure, and posture data into a structured, stakeholder-ready report in a few clicks.

Also Shipping in v3

Attack Chaining, the Exposure Command Center, the Adversarial Exposure Score, AI Red-Teaming, and automated reporting are the headline additions in this release, but they're not the whole story. v3 also introduces:

• Asset Criticality: letting teams tag assets with a criticality level tied to individual Asset Posture Scores, feeding prioritization across findings, expectations, dashboards, and the Command Center itself.
• New Communication Injectors: covering Email (both Microsoft 365 and Google Workspace), Slack, and Teams, extending social engineering and phishing simulation coverage to the channels attackers use.
• Microsoft Defender for Endpoint Executor: letting teams orchestrate scenario execution directly through Defender for Endpoint alongside the platform's growing list of XDR/EDR-driven executors, so detection and response can be validated without leaving the Microsoft security stack.
• Audit Logging: giving teams a full, traceable record of who did what across the platform, supporting the accountability and change-tracking that security and compliance teams need as more of the platform runs autonomously.
• A New Notifications Engine: providing clearer rules, more relevant alerts, and a foundation for richer notification types, following a ground-up rebuild.
• Multi-Tenancy Enhancements: enabling organizations to cleanly separate environments, teams, or customers within a single OpenAEV instance, making the platform practical to run at scale across business units or, for managed security service providers (MSSPs), across multiple client organizations.

A number of smaller improvements round out the release. For more information:

• Review the community release notes.
• Read the launch announcement.

Try OpenAEV v3

This release is ultimately about closing the gap between knowing your threats and proving your defenses hold up against them, continuously, automatically, and at a depth close to what a real adversary would actually throw at you.

• Start a free trial of OpenAEV v3.
• Join our upcoming webinar to see Attack Chaining and autonomous adversarial exposure validation in action.
• More on Attack Chaining: Check out my perspective piece, Attack Chaining: How to Test the Way Attackers Get In.