New Content & Platform Update
New Feature: Next-Gen AI Crisis Sim Builder
Release date: September 30th, 2026
We are launching the next-generation AI Crisis Sim Builder for Immersive One. The builder has been rebuilt as a conversational co-author that allows scenario authors to design, adapt, and refine full exercises through conversation rather than manual step-by-step editing. Authors can describe their scenario in plain text, upload up to five supporting files (such as incident response playbooks, threat intelligence reports, or post-incident reviews up to 50 MB each), and pull live threat detail via web search to automatically build tailored exercises.
The tool generates complete, ready-to-run exercises, including branching decisions with real dilemmas, multi-role seats (e.g., CISO, General Counsel, Head of Communications), dynamic feeds (emails, news, social media, and interactive AI character messages), exportable documents (PDF, DOCX, CSV), performance indicators, regulatory deadline clocks, and debriefs. Authors retain full control with live canvas updates, turn-by-turn change summaries, one-click undo capabilities, and automated scenario validation to prevent broken branches or dead ends. Additionally, the builder supports any authoring language, works with connected features like Threat Actor Insights and Goal Creation, and runs on the same underlying architecture as AI Lab Builder and AI Program Builder.
Who are we doing it for
This release is built for exercise owners, scenario authors, and content teams across organizations using Immersive One (Core and Advanced). It serves both technical and non-technical authors who need to build or adapt exercises for executive teams, specific industry sectors, or multi-role incident response drills without needing deep expertise in scenario-writing tools.
Why are we doing it now
Developing realistic crisis simulations traditionally requires significant time to craft coherent timelines, distinct roles, branching outcomes, and custom feeds. While our legacy AI builder offered one-shot generation, authors were forced to manually edit every subsequent detail, leading to frustration and reliance on manual workarounds. Furthermore, organizations increasingly require scenarios customized to their specific sectors, regulatory demands, playbooks, and recent security incidents rather than off-the-shelf templates. By moving to a unified conversational AI architecture, we enable rapid, flexible exercise creation and ensure continuous platform-wide improvements across all Immersive One builders.
New AI Knowledge Pass IdP Integration: Microsoft Entra
Release date: September 30th, 2026
AI Knowledge Pass now connects to Microsoft Entra, so customers can manage access to AI tools based on verified competency from within their existing Entra setup. No separate admin console required.
This option is available to Advanced customers.
Feature Update: Alias Search is now live in Threat Actor Insights
Release date: September 29th, 2026
We updated threat actor search in Threat Intelligence (TAI) to match every associated group name published by MITRE. You can now search using any alias such as "Charming Kitten", "Mint Sandstorm", "APT35", or "TA453" and TAI will automatically resolve the search to the primary entity, like Magic Hound. The search results explicitly show which alias matched your query, ensuring complete transparency on why a specific actor appears.
NEW CTI Lab - CVE-2026-87902 (WordPress Path Traversal)
Release date: September 29th, 2026
WordPress is an open-source Content Management System written in PHP and the most widely deployed CMS in the world. Organizations use it for everything from public-facing websites and e-commerce stores to internal portals and IT service desks, making it an attractive and consistently targeted platform. Its extensibility through themes and plugins means attack surface varies widely across deployments, and the sheer scale of its user base makes even narrow vulnerability windows significant.
An unauthenticated path traversal vulnerability (CVSS 8.1 High) was disclosed in WordPress core on September 22, 2026, affecting all versions from 4.7.0 through 7.1.1. The flaw allows an unauthenticated, remote attacker to bypass template path validation using double-encoded characters, causing WordPress to include arbitrary PHP files from the server. On common hosting configurations, this escalates to full remote code execution without any credentials. Exploitation was confirmed within hours of the patch release, with Patchstack and BleepingComputer both reporting active scanning and payload delivery shortly after WordPress 7.1.2 was published.
In this lab, users step into the attacker's seat. Taking on the role of a penetration tester conducting an authorized assessment of an internal IT services portal, they'll enumerate the target, identify the conditions that make it exploitable, and chain the path traversal into full remote code execution on the WordPress server.
Who are we doing it for
This lab is designed for penetration testers, red teamers, and security engineers who need to understand and demonstrate the practical impact of web application vulnerabilities in enterprise WordPress deployments.
It's also highly valuable for any cybersecurity practitioner who wants to understand how a seemingly limited path traversal flaw can escalate to full server compromise, and how pre-conditions in server configuration and theme structure turn a conditional vulnerability into a critical one.
Why are we doing it now
WordPress is the most widely deployed CMS on the internet, and CVE-2026-87902 requires no credentials, no plugin installation, and no prior foothold, just a vulnerable version and a server configuration that many deployments have in place by default.
We're delivering this lab now to give practitioners hands-on experience with the full exploitation chain, from reconnaissance through to shell access, so they can assess their own environments and understand exactly what exposure looks like.
Availability: This lab is available now, to all customers on all tiers.
Link to the lab: here
NEW CTI Labs - CVE-2026-7273 (Zyxel Buffer Overflow)
Release date: September 24th, 2026
A stack-based buffer overflow (CVSS 3.1 8.8) has been disclosed in the web management CGI of the Zyxel GS1900 series, patched by Zyxel on June 16, 2026. Since then, researchers have observed this vulnerability being exploited in the wild by threat actors. One threat group has been known to compromise 996 switches across 48 countries as part of a recent campaign. CISA subsequently added this CVE to the Known Exploited Vulnerabilities catalog on September 21, 2026.
In the offensive lab, users step into the attacker's seat. Taking on the role of an internal penetration tester, they'll confirm that a GS1900 switch is running a vulnerable firmware build, then prove the finding with real code execution on the device and recover the token stored on it.
In the defensive lab, users take on the role of a network security analyst responding to an alert. They'll work through a packet capture to identify the exploit, extract commands from the HTTP payload, and trace the follow-on activity.
Who are we doing it for
The offensive lab is designed for penetration testers and red teamers assessing network infrastructure, as well as the network and security engineers who own these devices and need to see exactly what a vulnerable switch hands to an attacker with adjacent access. The defensive lab is designed for security analysts, incident responders, and threat hunters who need to determine whether their network devices have been compromised, and to recognize the indicators of compromise this attack leaves in captured traffic.
Why are we doing it now
CVE-2026-7273 is under active exploitation. CISA's three-day federal remediation deadline reflects how seriously the community is treating the risk, but many deployments will not patch quickly.
We're delivering both labs now so offensive teams can confirm exposure in their own estates before attackers do, and blue teams can recognize known attack patterns in their captures and respond before credential harvesting or deeper network compromise follows.
Availability: These labs are available now, to all customers on all tiers.
New release: Securing the AI Supply Chain
Release date: September 23rd, 2026
We are launching Securing the AI Supply Chain, a hands-on collection serving as the practical counterpart to our recent Introduction to the AI Supply Chain. Across three interactive labs, teams get direct experience evaluating third-party machine learning models before they reach production. Learners scan model artifacts for malicious payloads, generate and read an AI Bill of Materials (AIBOM) to trace provenance, and put both controls to work in a real-world supply chain admission audit scenario.
Why are we doing it now
Supply chain attacks reached developer environments long before AI arrived, but what those pipelines now carry has changed fundamentally. Traditional AppSec tools and standard SBOMs are blind to poisoned neural network weights, serialized code execution, and hijacked registry namespaces. Attackers don't need complex zero-days when they can slip malware into trusted ML models on public hubs like Hugging Face or exploit flaws like CVE-2025-32434 to execute code. Securing an AI architecture requires stripping away the black-box mystique and enforcing concrete controls at the pipeline level.
Related content
This collection acts as the technical anchor for our AI defense curriculum. While Introduction to the AI Supply Chain builds the theory and our recent AI Supply Chain Crisis Sim tackles executive response, these new labs deliver the hands-on engineering controls needed to prevent incidents entirely. It drops perfectly into any AI Security Readiness or governance training track.
Availability: This new collection is live today and available to all customers across all tiers. Legacy SKUs: Immersive Labs & Application Security
Link to the collection: here
New release: System One Models : Jev
Release date: September 22nd, 2026
We are excited to launch a new theory lab called System One Models Jev. This lab explores a completely different breed of AI. These are not LLMs that write text; they are models designed for action. You feed them data, and they return a concrete decision your code can execute. The lab breaks down the core mechanics behind this technology, the three specific types of questions it can answer, and exactly when to use it over plain code, traditional classifiers, or LLMs. We have also published a companion blog on the website to support the release.
Why are we doing it now
Most AI security and governance conversations are entirely focused on generative models. But System One models introduce a different paradigm: automated, high-speed decision-making at scale. As organizations rush to integrate these models to drive automated workflows, security and engineering teams need to understand the mechanics behind those decisions. This lab strips away the generative hype and gives teams the technical grounding to know exactly when a System One model is the right tool for the job, and when traditional logic is safer.
What is next?
This theory lab is the starting line for a broader conversation on AI automation. It establishes the technical foundation that pairs perfectly with the companion blog live on our website today. More importantly, it serves as the essential prerequisite for a related crisis sim dropping next week. GTM teams can pitch this as a cohesive "AI Decision-Making" track: use the lab to understand the mechanics today, so your teams are ready to navigate the fallout in next week's simulation.
Availability: This lab is live today and available to all customers across all tiers.
Link to the lab: here.
New Crisis Simulation: AI Supply Chain - Monitoring Issues
Release date: September 18th, 2026
We are launching a new crisis simulation: AI Supply Chain: Monitoring Issues. Set on the eve of Black Friday at a multinational retailer, the scenario forces executive teams to manage a compromised third-party AI observability platform. Because the company routes all AI traffic through this single vendor, the platform holds a complete record of every prompt, completion, and pasted secret from tens of millions of customers and colleagues. When the platform is compromised, the crisis team must navigate a nightmare scenario: managing a live AI estate they can neither safely keep running nor lawfully switch off.
Why are we doing it now
AI adoption has quietly placed observability and monitoring platforms in front of the entire AI estate, concentrating massive amounts of sensitive data into single third-party stores. This makes them prime targets for mass-exploitation zero-days and data extortion crews—the exact pattern seen in incidents like MOVEit. Emerging regulations raise the stakes even further. Under the EU AI Act, logging for high-risk AI systems is legally required; you cannot simply turn the logging off because the vendor holding it has been breached. This scenario gives incident commanders, CISOs, and legal officers the chance to practice an event where operational continuity, regulatory duties, and customer trust collide, and there is no clean answer.
The AI Supply Chain: This simulation anchors our broader rollout of supply chain content. It links perfectly with the Introduction to AI Supply Chain (out now) and Securing the AI Supply Chain (coming next week) collections. Customers can use the new collections to build foundational skills and defensive capabilities within their technical teams, and then drop this crisis simulation in to pressure-test their cross-functional leadership.
Availability: Users must have Crisis Sim permissions to view the scenario link
Link to the crisis: here
New CTI Lab - CVE-2026-85706 (GitLab Path Traversal)
Release date: September 18th, 2026
GitLab is an open-source DevSecOps platform that combines source code management, CI/CD pipelines, issue tracking, and security scanning into a single application. It is widely deployed as a self-managed installation across enterprises, government organizations, and open-source projects, where it typically holds internal source code, infrastructure-as-code, deployment credentials, and release pipelines. This makes it a high-value target for anyone seeking access to an organization's software supply chain.
A critical path-traversal vulnerability (CVSS 10.0) has been disclosed in the GitLab CE/EE API, patched by GitLab on September 10, 2026. The flaw allows an unauthenticated, remote attacker to read arbitrary files from the server filesystem by sending a single crafted request. It exploits a parser difference between GitLab Workhorse and the underlying Ruby application, so an attacker-supplied file path reaches Rails and is read before any authorization check runs. Active exploitation has been confirmed, exposing secrets such as GitLab's encryption bundle, SSH host keys, and database credentials.
In this lab, users will step into the role of a SOC analyst investigating an incident. They'll work through raw nginx and GitLab API logs to establish what happened, identify the attacker, determine which files were targeted. They'll learn to recognise the Indicators of Compromise (IOCs) this attack leaves behind and how to detect and mitigate it in their own environments.
Who are we doing it for
This lab is designed for security analysts, incident responders, and threat hunters who monitor and defend developer and CI/CD infrastructure, as well as anyone responsible for log analysis and detection engineering. It's also highly valuable for any cybersecurity practitioner who wants to understand why platforms like GitLab are such attractive targets, and how a single parser inconsistency between layers of a stack can undermine an authorization boundary entirely.
Why are we doing it now
CVE-2026-85706 scores a maximum 10.0 on the CVSS scale, is under confirmed active exploitation, and requires no credentials - just one HTTP request to an internet-facing GitLab instance. A successful read hands attackers the secrets they need to move deeper into an organization's software supply chain, and public proof-of-concepts target multiple endpoints and encodings, so detections tuned to a single pattern will miss variants.
We're delivering this lab now to equip blue teams with the practical skill of turning raw access logs into detections, so they can recognise a path-traversal attack against GitLab in progress and respond before an attacker walks away with the crown jewels.
Availability: This lab is available now, to all customers on all tiers.
Link to the lab: here
CrowdStrike Next-Gen SIEM in Dynamic Threat Ranges
Release date: September 13th, 2026
Dynamic Threat Ranges can now connect to CrowdStrike and forward range telemetry into the platform. Teams can select CrowdStrike as their SIEM option, with each team’s data isolated from other teams.
This enables defenders to practise detection and investigation in the CrowdStrike environment their organisation uses.
Who are we doing it for
Blue team defenders and Security Team Leads
Why are we doing it now
Organisations that have standardised on CrowdStrike need training that reflects the tools their teams use in real security operations.
By bringing CrowdStrike into Dynamic Threat Range exercises, teams can:
- Build tool fluency: Investigate threats in the CrowdStrike environment used by the organisation.
- Connect training to readiness: Practise detection and investigation through a realistic blue team workflow.
- Reinforce the platform standard: Align security training with an organisation’s CrowdStrike investment and rollout.
Availability: Available now to Core and Advanced tiers customers.
New CTI Lab - BlueMoon Exploit Kit: Campaign Analysis
Release date: September 14th, 2026
In September 2026, security researchers at Proofpoint and Volexity announced they had uncovered a sophisticated browser exploit kit dubbed BlueMoon. This was able to silently compromise Windows systems through a single malicious web page, by chaining three zero-day vulnerabilities together. Two of the three bugs are patch-gap zero-days: vulnerabilities that had been fixed in the public Chromium source code but hadn't yet shipped to end users in a stable Chrome release. Researchers believed that BlueMoon's authors reverse-engineered the vulnerabilities directly from the public patch commits, moving from published fix to weaponised exploit in under three weeks.
Four separate threat groups, all believed to be aligned with China, deployed the kit against distinct targets within days of each other. The near-simultaneous use of identical code led researchers to assess the kit was sold or shared between operators rather than independently developed. All three CVEs (CVE-2026-85046, CVE-2026-87491, CVE-2026-85880) were subsequently added to CISA's Known Exploited Vulnerabilities catalog.
In this lab, users step into the role of a malware analyst examining recovered artefacts from a BlueMoon intrusion. Working through several files spanning the post-exploitation chain, they'll reverse engineer each component using static analysis to identify what it does, recover obfuscated configuration, and extract indicators of compromise.
Who are we doing it for
This lab is designed for malware analysts, threat hunters, and CTI practitioners who want hands-on experience analysing a real, multi-stage nation-state intrusion chain. It's also highly valuable for detection engineers and incident responders who need to understand the techniques before they can reliably hunt for them.
Why are we doing it now
The patch-gap technique used in BlueMoon is not new, but this campaign demonstrates it being operationalised faster and at greater scale than previously observed. The tradecraft is already in active use and will inform offensive operations well beyond this specific kit. Artefacts throughout the exploit code also suggest AI-assisted development contributed to the unusually short time from public patch to operational exploit, a trend that is accelerating the pace at which defenders need to respond to newly disclosed vulnerabilities.
We're delivering this lab now to give practitioners direct exposure to the analysis techniques needed to pull apart a modern, layered intrusion chain, so they can recognise the same tradecraft the next time it lands in their environment, under a different name.
Availability: This lab is available now, to all customers on all tiers.
Link to the lab: here.
New DTR Release: Nebula Bank - MITRE ATLAS
Release date: September 9th, 2026
Learners work through security tickets using Claude Code or OpenAI Codex. They reproduce reported issues, inspect and fix the source code, push their changes through Gitea CI, and validate the deployed fixes against security and functional tests.
The range covers a focused set of risks mapped to MITRE ATLAS and the OWASP Top 10 for LLM Applications, including:
- Prompt injection
- Unsafe AI tool use and excessive agency
- Broken authorization
- Sensitive data exposure
Who are we doing it for
- Software developers, especially those building AI applications
- Application security and product security teams
- AI security teams
Nebula expands our developer DTR portfolio alongside Blossom. While Blossom focuses on conventional web application security, Nebula applies the same hands-on remediation format to AI-enabled applications.
New Program: Red Teaming AI with MITRE ATLAS
Release date: September 9th, 2026
We are launching our new Red Teaming AI with MITRE ATLAS program on the Immersive One platform. This curriculum contains 19 hands-on activities that deliver 13 hours and 30 minutes of practical content structured across nine steps.
Learners will navigate three beginner, 14 intermediate, and two expert labs. You will learn how to systematically break and exploit live AI systems, including retrieval-augmented generation (RAG) pipelines and autonomous agents. The program ends with capstone assessments that validate skills across the full AI attack lifecycle, from Reconnaissance and Initial Access to Persistence, Lateral Movement, and Defense Evasion.
Who are we doing it for
It’s designed for red teams, penetration testers, and security engineers who need to build real-world offensive skills against modern AI-integrated infrastructure.
Why are we doing it now
Enterprise AI adoption is rapidly outpacing traditional security capabilities. Organizations are deploying AI agents and complex pipelines into production faster than their security teams can learn to test or attack them. This program provides a safe, interactive environment to functionally exploit AI systems instead of relying on theoretical risk models.
This release serves as the offensive foundation of our broader AI curriculum. It also represents the first half of a complete Red and Blue training path: as you learn how AI infrastructure is broken here, we are actively developing the defensive equivalent mapped to Elastic so your teams can detect these exact threats.
Availability: Live today for all customers with Programs and AI Lab access.
Link to the Program: here
New CTI Labs - CVE-2026-82329 (Artifactory Auth Bypass)
Release date: September 9th, 2026
We're releasing two new Cyber Threat Intelligence (CTI) labs covering CVE-2026-82329, one offensive and one defensive:
- CVE-2026-82329 (Artifactory Auth Bypass) – Offensive
- CVE-2026-82329 (Artifactory Auth Bypass) – Defensive
A critical authentication bypass (CVSS 3.1 9.8) has been disclosed in Artifactory, tracked as CVE-2026-82329. When the
additionalJoinKeys setting is left blank, Artifactory's Access service accepts a secret derived from the SHA-256 hash of an empty string. That value is fixed and publicly known, so an unauthenticated, remote attacker can forge a valid token from scratch, exchange it for a full administrative token, and read the entire global configuration. No credentials, no user interaction, and roughly four seconds from first request to admin.In the offensive lab, users step into the attacker's seat. Assuming the role of a penetration tester assessing their own infrastructure, they'll work through the full exploit chain: forging the join token, minting an admin token, and confirming the impact.
In the defensive lab, users take the SOC analyst's seat. Working from the raw Artifactory log files, they'll investigate a real intrusion: confirming the instance was vulnerable, separating the forged request from legitimate service traffic and finding the non-expiring admin token the attacker left behind.
Who are we doing it for
The offensive lab is designed for penetration testers and red teamers assessing DevOps and build infrastructure, as well as the platform, DevOps, and MLOps engineers who own these systems and need to see exactly what an exposed instance hands to an attacker.
The defensive lab is designed for security analysts, incident responders, and threat hunters who need to determine whether their systems have been compromised, and to recognize the indicators of compromise this attack leaves in application logs.
Together they're valuable for any cybersecurity practitioner who wants to understand why the software supply chain has become such an attractive target, and how a single blank configuration value can escalate to full administrative compromise.
Why are we doing it now
Active exploitation of CVE-2026-82329 has been confirmed. Artifactory is internet-facing in many deployments and holds the keys to the software supply chain, which makes it exactly the kind of high-value target attackers move on quickly once a critical, unauthenticated bug is public. The blank join key condition can appear without any deliberate action, so organizations may be exposed without realizing it.
We're delivering both labs now so offensive teams can identify exposure in their own estates and prove the real impact, while blue teams can recognize the attack in their logs and shut it down before an attacker escalates to full control of their build infrastructure.
Availability: This lab is available now, to all customers on all tiers.
New CTI Lab - CVE-2026-81578 (PaperCut)
Release date: September 4th, 2026
We're releasing a new Cyber Threat Intelligence (CTI) lab titled "CVE-2026-81578 (PaperCut) – Defensive".
PaperCut NG/MF is print management software used to track, control, and secure printing across an organization. It reaches over 100 million users in more than 70,000 organizations, including large enterprises, government agencies, and universities. A print server rarely looks like a crown jewel, but it's a trusted internal host that authenticates users, holds directory data, and runs as a privileged service account — exactly the kind of quiet foothold an attacker wants.
Two vulnerabilities have been disclosed in PaperCut NG/MF and chained together in real-world attacks: CVE-2026-81578 (CVSS 3.1 9.8), a missing-authentication flaw that lets an unauthenticated remote attacker reach a privileged configuration interface, and CVE-2026-82078 (CVSS 3.1 9.1), an unsafe-reflection flaw that turns those configuration writes into arbitrary code execution. Chained, they give a fully unauthenticated, remote compromise of the print server. CISA added both to the Known Exploited Vulnerabilities (KEV) catalog on August 31, 2026, and attackers and researchers were seen developing exploits within days of the disclosure, including Immersive Senior Director Cyber Threat Research, Kevin Breen.
In this lab, users won't exploit the vulnerability, instead they'll step into the defender's seat. Working in Elastic, they'll hunt the logs left behind by a PaperCut attack. They'll learn to recognise the Indicators of Compromise (IOCs) it produces, and understand how to detect and mitigate the technique in their own environments.
Who are we doing it for
This lab is designed for security analysts, incident responders, and threat hunters who monitor internet-adjacent application infrastructure, as well as anyone responsible for SIEM detection engineering.
It's also highly valuable for any cybersecurity practitioner who wants to understand why print management servers are such an attractive target, and how an unauthenticated configuration change can quietly escalate into code execution or bulk data theft.
Why are we doing it now
A CISA KEV listing means this vulnerability is being actively exploited in the wild, and PaperCut has a long history as a target for ransomware and state-linked groups. With attackers already exploiting this vulnerability, defenders are expected to catch it quickly.
We're delivering this lab now to equip blue teams with the practical skill of turning raw logs into detections, so they can recognise a PaperCut attack in progress and confirm whether an instance has been targeted and exploited before an attacker moves deeper into the estate.
Customer Availability: This lab is available now, to all customers on all tiers.
Link to the collection: here
New Release: AI Red Teaming — Defense Evasion
Release date: September 4th, 2026
- LLM Jailbreak (AML.T0054): Bypassing safety controls, content filters, and input/output guardrails using encoding, semantic substitution, and character spacing.
- False RAG Injection (AML.T0071): Exploiting the trust boundary between document retrieval and LLM inference, probing security layers from provenance validation to audit logging.
This release completes the puzzle for our full "Red Teaming AI with MITRE ATLAS" program launching next week. If your security teams are preparing to audit enterprise AI deployments, this collection serves as the final hands-on capstone to ensure your workforce can effectively evaluate and harden "safe" AI systems against adversarial inputs.
Customer Availability: available to all customers across all tiers. Please note that these labs contain live AI systems, and as such as unavailable to customers with no access to AI labs.
Link to the collection: here
Feature Update: AI Lab Builder
Release date: September 3rd, 2026
Smarter generation, file upload, an undo button, and a massively upgraded briefing editor.
We've shipped a big update to AI Lab Builder - sharpening the AI itself, giving it file upload and allowing authors to have far more control over how labs look and read.
This update improves our core artificial intelligence model to deliver higher-quality, more reliable lab content aligned with your intent, and introduces a built-in file upload feature. Authors can now upload files – such as security policies, frameworks, or binaries – so the AI can automatically build relevant content around real-world artifacts. We have also introduced a revert safety net to undo changes and easily step back to earlier versions.
In addition, we have overhauled the briefing editor with a rich formatting toolkit and pre-built templates. Authors and the AI can now structure briefings using horizontal rules, tables, cards, callout boxes, alert banners, and hover-definition acronym tags. These proven, learner-tested building blocks standardise sections like pre-requisites, detection indicators, and real-world examples across your catalogue.
Note: While the AI can generate end-to-end theory labs and edit all non-machine elements, it cannot currently create or edit practical machines.
Who are we doing it for
This release is for all content authors, instructional designers, and subject matter experts (SMEs) building custom learning content on the Immersive One platform. It also benefits junior learners and cybersecurity professionals taking these labs, as well as platform administrators looking to reduce user support tickets.
Licensing remains unchanged, and the enhanced briefing components are available to all customers immediately. To access the AI assistant features, administrators simply need to enable AI in their platform settings.
Why are we doing it now
We are releasing this now to transform lab authoring from time-consuming, freeform formatting into assembling consistent, high-quality content quickly. Custom content creation often presents formatting challenges, leading to plain walls of text or visual inconsistencies.
New collection release: Introduction to the AI Supply Chain
Release date: September 3rd, 2026
We have launched the first collection in our AI pipeline series: “Introduction to the AI Supply Chain”. This collection breaks down the four primary architectural layers of modern artificial intelligence (AI) implementations and highlights the hidden security vulnerabilities within each layer. You will explore the evolving threat landscape and gain practical knowledge of key security controls necessary to protect your machine learning (ML) environments. This foundational collection serves as the entry point for our broader AI defense curriculum and will be followed by the “Securing the Supply Chain” collection, featuring practical labs and supporting Crisis Simulation content.
Why are we doing it now
Software supply chain attacks targeting developer environments and continuous integration and continuous delivery (CI/CD) pipelines are an established threat, but the assets moving through these pipelines have changed drastically with the rise of AI. Traditional AppSec tools and standard software bill of materials (SBOM) frameworks are blind to poisonings in neural network weights, serialized code execution, and hijacked registry namespaces. Threat actors no longer need complex zero-day exploits; they can simply slip malicious code into trusted ML models or developer tools to bypass perimeter defenses. We are releasing this collection now to help you strip away the “black box” mystique of AI pipeline vulnerabilities and implement concrete, actionable security controls across your architecture.
Who are we doing it for
This feature is designed for application security (AppSec) engineers, security architects, software developers, and technical cybersecurity teams who build, deploy, or secure modern AI architectures and software development pipelines.
Link to the collection: here
Availability: Core license
New collection release: OWASP Agentic Skills Top 10 (2026)
Release date: September 3rd, 2026
We are releasing a brand-new theory collection covering the complete OWASP Agentic Skills Top 10. This foundational track breaks down all ten risk categories—from Critical threats like Malicious Skills (AST01) and Supply Chain Compromises (AST02) to operational risks like Update Drift (AST07) and Cross-Platform Reuse (AST10). While this is a knowledge-based collection rather than a hands-on lab, it provides the exact framework teams need to understand, classify, and mitigate vulnerabilities in AI agent ecosystems.
Why are we doing it now
As autonomous AI agents are integrated into enterprise workflows, they introduce a dual-layer attack surface. Unlike traditional software packages, agentic skills can be exploited through both standard code execution and the natural language instruction layer (like markdown prose). With active threat campaigns already successfully publishing malicious skills to public registries to steal credentials and inject backdoors, traditional application security scanners are largely blind to these new vectors.
This collection acts as the perfect prerequisite to our hands-on AI and LLM security content. It is an excellent technical module to drop straight into your AI governance, secure AI development, or threat modeling training paths, giving your teams the necessary vocabulary and theoretical foundation before they tackle our practical AI exploitation labs.
Who is this for
Governance, AppSec, and AI development teams need a structural understanding of this threat landscape before they begin deploying or approving agent frameworks.
Link to the collection: here
Availability: This collection is available to all customers on all tiers, and is available now!