Archive for March, 2026

Cobalt Introduces New AI Capabilities for Continuous Pentesting

Posted in Commentary with tags on March 19, 2026 by itnerd

Cobalt today announced new AI capabilities for continuous pentesting. Delivered through the Cobalt Offensive Security Platform, these next-generation components integrate AI with elite human pentesters and more than a decade of proprietary pentesting intelligence to accelerate the speed, scale, and depth of modern offensive security programs. Attendees of the RSA Conference can learn more by visiting the Cobalt team at Booth #N4519 at the Moscone Conference Center.

Offensive security is entering a new era. Attackers are increasingly using AI to automate reconnaissance, vulnerability discovery, and exploitation. At the same time, modern development practices are accelerating release velocity and dramatically expanding the attack surface across APIs, microservices, cloud infrastructure, and AI-powered applications. Security teams can no longer rely on periodic testing to understand their exposure—they must validate real-world risk continuously.

The Cobalt Platform enables organizations to move beyond point-in-time testing and adopt a programmatic approach to offensive security that continuously adapts to evolving environments. Using the largest dataset of real-world pentesting intelligence in the industry, it applies historical exploit intelligence to refine testing logic and ensure every engagement is smarter than the last. Cobalt integrates and exposes the industry’s most capable hacker tools—constantly updated to reflect current threat actor tactics.

New features and functionality include:

  • Automated Reconnaissance: The AI-powered platform autonomously maps the entire attack surface—from complex JavaScript routes to hidden shadow APIs and forgotten subdomains. This identifies every potential entry point and provides human testers with a high-fidelity roadmap from the start of every engagement. 
  • AI-Powered Vulnerability Discovery: By combining automated scanning with AI-driven credential validation, the Cobalt Platform ensures exhaustive coverage of all form fields and CVEs, including critical vulnerabilities like those in Log4j and WordPress. This autonomously validates access and surface-level flaws to provide an immediate baseline of enterprise risk.
  • Proprietary Data Enrichment: Every finding is enriched with context from public exploit feeds and over a decade of proprietary historical intelligence. By merging global threat data with a unique offensive security dataset, the Cobalt Platform provides the critical context needed to frame findings based on actual adversarial behavior.
  • AI-Driven Deduplication and Triage: An AI-driven triage engine automatically normalizes and deduplicates findings across all scanner outputs into a single, cohesive view. By distilling high-volume data into verified findings, the platform ensures pentesters are focused on creative attack scenarios that present the real risk to the business.

These enhancements build on additional AI capabilities released in Q4 2025, including AI-Powered Reporting and Insights. AI reporting automates vulnerability documentation, benchmark results against aggregated security data, and provide natural-language access to product guidance. By combining an AI report writer, insights and benchmarking capabilities, and an AI documentation assistant, the Cobalt Platform accelerates report delivery, contextualizes findings with industry data, and helps security teams quickly understand and remediate risk.

With only a few clicks to scope and set up a pentest, the Cobalt Platform initiates testing automatically to ensure depth and quality before human experts engage. Because reconnaissance and scanning are now fully automated, pentesters spend 0% of their time on basic discovery and 100% of their time on high-value exploitation. 

The Cobalt Platform also introduces compatibility with the Model Context Protocol (MCP), enabling AI assistants to securely interface with pentest data so security teams can query testing results, triage findings, and correlate risk through natural-language workflows. 

Additional Resources: 

Cobalt Introduces Security Program Manager Service to Help Enterprises Scale Offensive Security Programs

Posted in Commentary with tags on March 19, 2026 by itnerd

Cobalt today announced the launch of its Security Program Manager service, designed to help enterprises operationalize and scale offensive security programs. Attendees of the RSA Conference can learn more about these new capabilities by visiting the Cobalt team at Booth #N4519 at the Moscone Conference Center.

As organizations expand their security testing efforts across applications, APIs, cloud infrastructure, and emerging technologies, many security teams struggle with a growing gap between strategic security objectives and day-to-day execution. Fragmented oversight, engineering silos, and the challenge of translating technical vulnerability data into business-level insights can slow remediation efforts and reduce the effectiveness of offensive security programs.

The Cobalt Security Program Manager addresses this challenge by providing organizations with a dedicated expert who acts as an extension of the internal security team. Security Program Managers oversee the logistics of enterprise-scale pentesting programs, coordinate testing schedules across development teams, and ensure remediation workflows align with broader business and security goals.

Security Program Managers help organizations streamline pentesting operations and ensure testing results translate into actionable improvements across the business. Key benefits of the service include:

  • Reclaim Your Team’s Time: Security Program Managers coordinate with development and engineering teams to schedule pentests, manage administrative logistics, and track remediation progress, reducing the operational burden on internal security teams.
  • Eliminate Security Blind Spots: By maintaining a comprehensive inventory of assets and aligning testing cadences with corporate security objectives, Security Program Managers ensure continuous visibility into the organization’s security posture.
  • Secure Executive Buy-In: Security Program Managers translate technical findings into strategic intelligence and performance metrics, helping security leaders demonstrate ROI and communicate risk reduction to executive stakeholders.
  • Accelerate Innovation Cycles: Cobalt integrates pentesting workflows with common development tools such as Jira, GitHub, and Slack, enabling organizations to embed security into development pipelines without disrupting engineering velocity.

The Security Program Manager builds on the broader Cobalt Offensive Security Platform, which combines automation, AI-driven intelligence, and expert-led testing to deliver offensive security at enterprise scale. By integrating automated reconnaissance, vulnerability discovery, and intelligence-driven triage with human-led testing, Cobalt enables organizations to run continuous security programs that evolve alongside their environments.

Cobalt offensive security services span application, network, API, cloud, and emerging AI systems, and include capabilities such as web application pentesting, mobile testing, cloud configuration reviews, attack surface management, red teaming, and AI and LLM application testing. These services are delivered by the Cobalt Core, a global community of more than 500 vetted ethical hackers who average over 11 years of pentesting experience.

Bonfy Unveils First Data Security Platform for AI Agents, Shadow AI, and Enterprise GenAI Workflows

Posted in Commentary with tags on March 19, 2026 by itnerd

Bonfy.AI today announced Bonfy Adaptive Content Security™ (Bonfy ACS) 2.0, the industry’s first platform built to secure enterprise content across all systems, applications, and AI agents – anywhere data moves, resides, or is processed. As organizations race to deploy copilots, custom AI apps, and increasingly autonomous AI agents, security leaders are struggling with blind spots around how these systems access, transform, and share sensitive data, gaps that legacy DLP and DSPM tools were never designed to handle. By 2028, Gartner projects that 22% of cyberattacks and data leaks will involve generative AI, and through 2029 over 50% of successful cybersecurity attacks against AI agents will exploit access‑control issues.

Bonfy delivers real-time, contextual protection across email, SaaS apps, collaboration tools, browsers, cloud and on‑prem file stores, AI systems, and agent frameworks, so enterprises can safely accelerate AI adoption without flying blind. With native coverage for Microsoft 365 (Exchange Online, SharePoint, Entra, Copilot and Purview), Google Workspace (Gmail, Google Drive, Google Directory), Salesforce, HubSpot, Slack, on‑premises file stores, AWS S3, and more, Bonfy becomes the unifying data security layer that follows content regardless of channel or AI workflow. 

Built for agentic and autonomous AI

Bonfy ACS 2.0 is engineered specifically for system‑level and browser‑based AI agents that plan, reason, call tools, and execute actions across enterprise systems.

By treating agents as first‑class entities, not just extensions of users, Bonfy allows security teams to see which agents accessed which data, how they used it, and where the outputs ultimately landed. Bonfy ACS 2.0 fits cleanly into a customer’s existing security and productivity stack: it complements Microsoft Purview and M365 DLP, integrates with Microsoft Entra and Google Directory, and plugs into SIEM/SOAR tools such as Splunk, Sentinel, and Rapid7 for workflow automation, while also integrating via its MCP Server interface and APIs with modern AI platforms including Microsoft Copilot Studio, OpenAI, Anthropic Claude, Google Gemini, and other enterprise agent frameworks.

Bonfy is designed for executive level visibility and governance for CISOs, CIOs, and to be operated by security teams, security architects, and AI platform teams responsible for GenAI and agent deployments in financial services, insurance, technology, biotech/pharmaceutical, healthcare companies, and more. It provides one policy and automation engine that spans traditional data security, AI data governance, and AI agent guardrails, eliminating the need to stitch together separate point products for systems, and agents. The Bonfy platform can now be used both for projects where organizations consume AI and build AI.

Headline capabilities in Bonfy ACS 2.0

Bonfy ACS 2.0 introduces six major capabilities that together form a second-generation, high‑performance data security platform for the AI era.

  • AI Agent Data Guardrails (MCP & Agent Framework Support)
    Bonfy adds “data in use” security solution by adding an MCP server interface, API, and agent‑aware controls so enterprises can inspect and govern the content AI agents read, share, and generate during planning, reasoning, and execution, not just in the final output. Agents can call Bonfy inline to label and risk‑score content before it reaches external services or users, stopping AI‑driven leakage and trust‑boundary violations.
  • Browser Extension for Shadow AI and Agentic Activity
    A lightweight browser extension delivers real‑time, content‑aware inspection of web traffic, including unsanctioned AI tools and browser‑based assistants. Bonfy separates safe AI use from risky disclosure, detects shadow AI automations, and shows security teams exactly where sensitive data is going.
  • Full Google Workspace Support
    Bonfy 2.0 adds native support for Gmail, Google Drive, and Google Directory, achieving parity with Microsoft 365 integrations and extending multi‑channel protection across both ecosystems. Organizations running on Google now get unified, entity‑aware controls including contextual, automated classification labeling.
  • Data Surface Visibility for AI-Era Risk
    A new “data surface visibility” view gives CISOs a live map of where sensitive content lives across data stores such as Microsoft SharePoint, Google Drive, AWS S3 buckets, On-prem file stores and AI systems, and how employees and agents use it. Teams can drill from high‑level exposure down to specific actors and flows to understand real business risk, not just isolated events.
  • On-Premises and Cloud File Store Coverage
    Bonfy now covers on‑premises file stores and cloud object storage such as AWS S3, alongside existing SharePoint, Google Drive and other SaaS applications. This creates a unified control plane for unstructured data at rest, in motion, and in use.
  • Data Minimization, Encryption Enhancements, and SOC 2
    Bonfy 2.0 tightens data minimization, encryption, and configurable retention so the platform itself has a smaller, better‑protected footprint. Completing SOC 2 Type 2 certification as part of the release reinforces Bonfy’s readiness for highly regulated industries.

Availability and RSAC 2026

Bonfy ACS 2.0 is available immediately. RSAC 2026 attendees can schedule a live demo by contacting Vishnu Varma

“DarkSword” iOS Exploit Can Steal Data from iPhones

Posted in Commentary with tags on March 18, 2026 by itnerd

Researchers have uncovered a new iOS devices exploit kit dubbed “DarkSword” used to steal data from potentially millions of iPhones running iOS 18.4 through 18.6.2. The attack is linked to the Russian hacking group UNC6353 which recently used the Coruna exploit chain reported by Google and iVerify

Brian Bell, CEO of customer identity and access management platform FusionAuth, provided the following comments:

“When a device can be silently compromised when visiting a website, perimeter-based and device-based security collapse. That’s not a future risk, it’s the current reality for anyone with a mobile user base.

The right response isn’t to wait for your users to patch. It’s to build authentication that assumes the device is already compromised. Short-lived tokens, step-up authentication before sensitive actions, forced re-authentication when signals change. Design for the breach, not against it.

And here’s the piece that most teams miss: most authentication platforms are SaaS; your token policies, session controls, and audit logs live in someone else’s cloud, under someone else’s access controls. But when authentication runs inside your own infrastructure, isolated from external dependencies, a compromised device doesn’t cascade into a compromised system. Identity is your last defense, so make sure you own it.”

If you are worried about this new exploit, the fix is simple. Which is to update to iOS 26 as that apparently is not affected. The most recent version of iOS 18 which at the time of this article is 18.7.3 is also not affected. But I would just go straight to iOS 26 as it is likely to protect you from more than this single exploit.

SOCRadar Launches Redesigned VAR Program And Appoints Cybersecurity Partnership Leader Brian Costello as VP of Global Partnership

Posted in Commentary on March 18, 2026 by itnerd

SOCRadar today launched a new  “partner first” VAR program that will help its partners gain a greater competitive edge and increased profitability leveraging the company’s innovative AI-driven Threat Intelligence enriched with External Attack Surface Management, Digital Risk Protection, Marketing Resources, supported by a dedicated channel partner team.​ The program introduces expanded incentives, enhanced sales enablement, and deeper technical engagement designed to help high-performing partners accelerate pipeline and deliver greater value to customers.

Concurrent with the launch of its new VAR program, SOCRadar has appointed Brian Costello as VP of Global Partnerships. Throughout his career, Brian has built successful channel partner programs and led high-performing security and cloud teams, delivering innovative technology solutions that exceed targets and drive consistent year-over-year business growth in both Fortune 15 and emerging tech company spaces.

Aggressive Discounts, Stronger Partner Margins, Financial Incentives and More

The new VAR program will enable partners to benefit from more aggressive discounts and performance-based rewards, creating stronger margins and clear financial incentives for driving new opportunities. The program also introduces enhanced deal protection, incumbency advantages, and robust deal registration, ensuring partners are recognized and rewarded for the opportunities they develop.

To help partners engage customers more effectively, SOCRadar is also providing sales-focused enablement content, including concise use cases, problem statements, qualifying questions, and buyer personas. These resources enable partners to quickly identify opportunities and initiate meaningful security conversations without requiring an immediate deep technical dive.

The new VAR program includes:

  • Expanded and more aggressive discounts and rewards for high-performance partners accompanied by deeper sales and technical support.
  • Stronger margins – Discounts, incumbency and deal registration protection
  • Sales focused content – Easy to consume use cases, problem statements, qualifying questions and buyer personas to accelerate opportunity discovery without having to do a deep technical dive
  • Technical support – Scoping support, POV access and integrations mapping to understand how to fit into a customer’s environment
  • Event support – Funding and coordination to drive opportunities with prospects and customers in the field and build a close collaboration between the SOCRadar team and its partners.

SOCRadar is also expanding technical collaboration and support, offering partners scoping assistance, proof-of-value (POV) access, and integration mapping to demonstrate how the platform fits seamlessly within a customer’s security environment. In addition, the company is investing in joint field engagement, providing event funding and coordinated support to help partners generate demand, build pipeline, and strengthen relationships with prospects and customers. This initiative reinforces SOCRadar’s commitment to building close, high-impact partnerships that drive shared success in the market.

Oleria Ends the Era of Legacy IGA with the Launch of Adaptive Identity Governance

Posted in Commentary with tags on March 18, 2026 by itnerd

Oleria today announced Adaptive Identity Governance, a fundamentally new approach to identity governance built on its data-first, AI-native platform. Deployable in under an hour and fueled by rich identity intelligence, Adaptive Identity Governance finally gives organizations clear insight into who has access to what and whether that access is justified.

At the core of Adaptive Identity Governance is Oleria’s Trustfusion platform, which continuously aggregates identity, entitlement, and activity data across cloud, SaaS, on-premises, and custom applications. This unified identity intelligence layer enables adaptive governance and empowers  scale-ups to Fortune 10 organizations to make access decisions based on real-time identity and activity data rather than siloed tools, static role models, or infrequent review cycles.

With Oleria, organizations can govern and control access across the most complex enterprise environments, with a system that adapts alongside growing SaaS adoption, machine identities, and AI agents. Customers are empowered to:

  • Make defensible access decisions with real activity and peer insights
  • Continuously enforce least privilege with automated lifecycle governance
  • Enable secure, fast through service workflows embedded in employee productivity tools

Learn more here about how enterprises are replacing legacy IGA with adaptive identity governance or come find Oleria at the RSAC 2026 Conference, Booth #5164, North Hall.

Posted in Commentary with tags on March 18, 2026 by itnerd

Cybersecurity researchers at Akamai are reporting a sharp rise in malicious online activity following the outbreak of conflict involving Iran, with cybercrime increasing by 245% since late February. The surge includes widespread activity such as credential harvesting attempts, automated reconnaissance, and probing of enterprise infrastructure as attackers capitalize on geopolitical instability.

The financial sector has been the most heavily impacted, accounting for approximately 40% of observed malicious traffic, followed by e-commerce, gaming, and technology companies.

Researchers also observed significant increases in:

  • Automated reconnaissance traffic – Up 65%
  • Credential harvesting attempts – Up 45%
  • Infrastructure scanning for exposed services Up 52%
  • Botnet-driven discovery traffic – Up 70%
  • DDoS reconnaissance – Up 38%

Analysts warn that the volume and sophistication of activity are likely to persist as cyber operations continue to accompany broader geopolitical tensions.

Sunil Gottumukkala, CEO, Averlon provided this comment:

   “The surge in activity following geopolitical tensions is consistent with what we typically see in these environments. Early-stage signals like reconnaissance, credential harvesting, and infrastructure probing tend to increase significantly as attackers look for initial access opportunities.

   “Enterprises should assume this activity will persist and focus on preparedness. That means staying on top of attack surface and exposure management to reduce exploitable vulnerabilities and ensure known weaknesses cannot be used to gain initial access. It also means strengthening identity security and monitoring for credential misuse, since many of these campaigns rely on stolen credentials.

   “The organizations that fare best are the ones that treat this activity as a precursor to more targeted attacks and invest in visibility into their exposure and rapid remediation of high-risk issues.”

Michael Bell, Founder & CEO, Suzu Labs supplied this comment:

   “The 245% number is real but the breakdown underneath it matters more than the headline. Only 14% of the malicious traffic Akamai observed originated from Iranian IPs. Russia accounted for 35% and China 28%, which tells you this isn’t just Iranian retaliation. Russia and China are taking a “never let a good crisis go to waste” approach, using the conflict as operational cover to ramp up scanning, credential harvesting, and infrastructure mapping while defenders are focused on the named adversary.

   “The attack mix confirms it. Botnet discovery traffic up 70% and automated reconnaissance up 65% means most of what Akamai is measuring is the setup phase, not the main event. The actual attacks that follow this reconnaissance, using the access and mapping being built right now, will be worse than the current numbers suggest.”

Phillip Wylie, Chief Security Evangelist & Senior Consultant, Suzu Labs follows with this comment:

   “Geopolitical conflict has always created opportunity for cyber threat actors, whether they are nation-state aligned groups, cybercriminals exploiting distraction, or opportunistic attackers taking advantage of heightened uncertainty. What we are seeing now is consistent with historical patterns where global instability increases scanning, credential attacks, and reconnaissance activity as organizations shift attention toward crisis response.

   “What stands out is not just the volume increase but the automation behind it. Attackers are clearly leveraging AI-assisted tooling, botnets, and automated discovery techniques to quickly identify weak points while defenders are distracted. This reinforces the importance of continuous exposure management, strong identity security, and monitoring for abnormal reconnaissance behavior, not just traditional alert-driven detection.

   “Organizations should treat these spikes as a reminder that external events often translate into increased cyber risk. Security teams should prioritize basic defensive discipline such as patching exposed services, enforcing MFA, monitoring for credential abuse, and validating DDoS readiness. In periods of global tension, good cyber hygiene and visibility often make the biggest difference.”

Jacob Warner, Director of IT, Xcape, Inc. had this comment:

   “The recent surge in Iranian cyber activity following Operation Epic Fury highlights a sophisticated “loud vs. quiet” strategic pivot. High-profile “wiper” attacks, where large amounts of data are deleted, on entities like Stryker dominate headlines and cause immediate operational paralysis. Meanwhile, state-sponsored actors are simultaneously executing quiet, long-term espionage campaigns.

   “For security professionals, the danger lies in the “loud” attacks serving as a massive smoke screen, drawing incident response resources away from deep-seated persistence in critical infrastructure.

   “Defenders must look past the immediate carnage of defacements and wipers to hunt for “living off the land” techniques and compromised administrative tools like UEM and MDM platforms. Prioritizing identity security and behavioral analytics is the only way to catch the quiet intruder while the sirens are blaring.

   “In modern conflict, the wiper attack is just a loud invitation to a heist that has been running for months.”

We clearly live in interesting times. That is a bad thing at the moment as threats from threat actors are all around us. Meaning that we all have to be on our toes to counter those threats.

Review: Sharp Dynabook Portege Z40L-N

Posted in Products with tags on March 18, 2026 by itnerd

This is the second of two Sharp Dynabook laptops that I am reviewing this week. You can read about the firstSharp Dynabook that I reviewed here. And that was a light and reasonably quick laptop. Today I have something that is lighter than that one and faster in a lot of ways. It’s the Portege Z40L-N and it clocks in at a mere 2.11 pounds which is insanely light. In fact, if I needed a notebook to carry on the daily, this would likely be my choice.

But just because it’s light doesn’t mean that you give up anything. I’m going to start with performance based on the following specs:

  • Intel Core Ultra 7 Processor 258V
  • Windows 11 Pro
  • 14.0″ diagonal widescreen
  • 32 GB RAM
  • 512 GB PCIe NVMe SSD
  • Fingerprint reader
  • 2 x USB-C/Thunderbolt 3 ports
  • Intel ARC Graphics
  • WiFi 7, Bluetooth, Ethernet
  • 56 W battery

First let’s start with the CPU and GPU performance. I did my testing using Geekbench 6 and I did one test on battery and one test on AC power as PC laptops can have radically different results in each scenario. Here’s the results:

On battery:

  • Single Core: 2144
  • Multi Core: 9292
  • GPU (OpenCL): 26777

And here’s the results while on AC Power: 

  • Single Core: 2544
  • Multi Core: 9784
  • GPU (OpenCL): 24840

To put that in perspective, my M1 Pro MacBook Pro hit these numbers (both on battery and on AC power) for the CPU:

  • Single Core: 1762
  • Multi Core: 12431

So despite being a small and thin laptop, it put out pretty impressive numbers. And I had to run the GPU test three times to validate the numbers as it was odd that it had a higher score on battery versus on AC power. But the results came out the same every time so I have to assume that this is a legitimate score.

Now onto the disk test. Here’s the results on both AC and battery power:

  • Read: 3420.61 MB/s
  • Write: 2319.43 MB/s

So it’s not the fastest SSD around as my MacBook Pro posted up a score of over 5000 MB/s in both read and write tests. But it’s far from slow.

You also aren’t giving up anything in terms of ports:

The screen itself is very bright and vibrant. Though it does have issues with really bright reflections such as a camera flash. I won’t hold that against Sharp. The keyboard has a nice typing feel to it as well. I wasn’t a fan of the trackpad as using a MacBook Pro as my daily driver, I found getting used to a mechanical trackpad to be a problem as it is a mechanical diving board trackpad. But that’s a me problem. I am pretty sure it will not be a you problem.

On the left side, you get a USB-C/Thunderbolt 3 port for power and data, an HDMI port, a USB-A port, a headphone jack, and a microSD slot.

On the right side you get a USB-A port, a USB-C/Thunderbolt 3 port and an Ethernet jack along with an Kensington lock slot.

In short, This notebook is fully featured and missing nothing. Except maybe for battery life. It’s rated for “up to” 8 hours. I barely got past 5 hours. Your mileage may vary on that front. But given the overall package that’s on offer here, I will give Sharp a free pass on that one. You can pick one up for about $2600. And in my opinion, if you want a light laptop that doesn’t force you to give up on speed or connectivity, it would be money well spent.

Windsurf IDE Extension Drops Malware via Solana Blockchain Targeting Developers In The Process

Posted in Commentary with tags on March 18, 2026 by itnerd

Bitdefender has released research warning of an active attack using a malicious extension for the Windsurf IDE (integrated development environment). The campaign intentionally targets software developers, who typically have privileged access, API keys, and other high-value credentials.

Disguised as a legitimate R programming language tool, the extension installs a multi-stage NodeJS credential stealer that retrieves encrypted payloads from the Solana blockchain, leveraging legitimate third-party infrastructure instead of traditional command-and-control (C2) servers to evade detection.

Cybercriminals are increasingly abusing trusted developer ecosystems and decentralized infrastructure to plant malware and establish persistence.

You can read the research here: https://www.bitdefender.com/en-us/blog/labs/windsurf-extension-malware-solana

TrojAI Extends Enterprise AI Security with Agent-Led Red Teaming, Runtime Intelligence, and Coding Agent Protection

Posted in Commentary with tags on March 18, 2026 by itnerd

TrojAI today announced major new capabilities designed to secure the growing deployment of agentic AI in the enterprise going beyond the prompt layer. 

Agent-Led AI Red Teaming

TrojAI Detect now includes Agent-Led AI Red Teaming,which uses coordinated autonomous agents to conduct red team testing on AI agents, applications and models. This advancement allows AI security teams to easily perform complex testing scenarios that map to a wide range of known security frameworks with the click of a button. 

Key features include: 

  • Agentic testing: Specialized agents work together to test AI models, apps and agents, automatically correlating results into a single, actionable report.
  • Multi-turn attacks: Agents automatically orchestrate multi-turn and dynamic attack chains, eliminating manual configuration and using TrojAI’s vast library of datasets and manipulations.
  • Adaptive learning: Testing agents retain history and memory to evolve strategies across attacks, becoming more effective with each new cycle of testing.
  • Framework mapping: Test results are automatically mapped to OWASP, MITRE and NIST. 

Agent-Led AI Red Teaming transforms AI security testing from a complex, multi-step process into a streamlined, intelligent assessment aligned to industry-standard frameworks.

To learn more about how TrojAI secures AI through Agent-Led AI Red Teaming, read the full blog.

Agent Runtime Intelligence

To complement build-time risk assessment, Agent Runtime Intelligence is available as a new platform capability in private preview. It goes beyond the prompt layer to capture and analyze full AI agent execution traces, giving enterprises deep visibility into how AI agents behave at runtime, including tool usage, memory access, data retrieval patterns and system prompt exposure. This enables security teams to govern, test and enforce policy across complex, multi-step AI workflows.

With Agent Runtime Intelligence, TrojAI enables visibility for: 

  • Tool exposure and excessive agency
  • Prompt injection propagation across workflows 
  • Sensitive data access during retrieval 
  • System prompt exposure and memory interactions

The capability integrates seamlessly with TrojAI’s existing dashboards, MCP governance, SIEM integrations and compliance tooling.

Real-Time Protection of Coding Agents

As AI coding agents become embedded in development workflows, they introduce a new class of security risk. Real-Time Protection of Coding Agents extends TrojAI Defend to safeguard AI coding assistants such as Claude Code and Codex as they generate, retrieve and modify code.

The capability detects exposed secrets, prevents sensitive data leakage, including PII, and blocks indirect prompt injection attacks, such as malicious instructions embedded within a retrieved file. By monitoring agent behavior in real time, TrojAI ensures that coding agents operate within defined security guardrails without disrupting developer productivity.

With these three platform enhancements, TrojAI is redefining how enterprises protect the next generation of intelligent systems so they can confidently embrace AI innovation securely, transparently, and at scale.