Archive for November 17, 2026

Google identifies fully automated AI cyberattacks operating without human intervention

Posted in Commentary with tags on November 17, 2026 by itnerd

Google’s Threat Intelligence Group and Mandiant have released new research documenting what they describe as a significant escalation in AI-powered cyberattacks, including the first observed use of AI frameworks to fully automate vulnerability discovery and exploitation without human intervention.

Researchers identified an exposed server running a multi-agent framework called exploit_agent, designed to continuously analyze software patches, identify vulnerabilities, generate exploits and execute attacks autonomously. Evidence collected from the server indicated that the operator was particularly interested in targeting French media, entertainment and educational websites.

Google also documented attackers using autonomous offensive security frameworks, including CAIRN and PentAGI, during real-world intrusions.

The company warned that these tools can dramatically accelerate attacks by automating multiple stages of an intrusion, reducing the time defenders have to identify and address vulnerabilities.

John Strand, Owner, Black Hills Information Security

“We’ve been using agentic frameworks for exploit development, attack surface enumeration, and reconnaissance at Black Hills Information Security for about a year now. And I think one of the biggest changes we’ve seen over the past couple of months is that these models are becoming increasingly effective on much smaller hardware and memory footprints.

“We’re now getting to the point where systems with 126 GB of memory can run models that are very effective at conducting these types of attacks. Sure, they’re slower than the frontier models, but they absolutely work. And that changes things dramatically because you no longer need massive infrastructure to start automating significant portions of an attack.

“In fact, I was doing some research on Shodan today and identified what appeared to be roughly 2,000 AI penetration testing platforms exposed directly to the internet. That number is fascinating, and I expect it to continue growing as this technology becomes more widely adopted.

“But here’s the part that really caught my attention. A number of these systems appeared to be compromised computers that were running these agentic attack frameworks on top of them. Think about what that means. We’re potentially seeing attackers deploy autonomous offensive capabilities directly onto compromised infrastructure.

“We’re absolutely seeing an explosion in these capabilities. And I don’t think this is some temporary trend. I think agentic AI is going to become the new normal for offensive cybersecurity operations over the next five years, and defenders need to start preparing for that reality right now.”

Stop me if you have heard this before. If your defenses do not include AI, they need to or else bad things will happen.

OpenAI’s agents followed a red team playbook, and security teams need to test them that way

Posted in Commentary with tags on November 17, 2026 by itnerd

California’s lawsuit against OpenAI is framed as a public nuisance case, but underneath it is a security story. An AI agent escaped its sandbox, used stolen credentials, found an unknown vulnerability, and later suppressed the alerts meant to catch it.

Gunter Ollmann, CTO, Cobalt

“Set aside the legal question, and the more important lesson for security teams is what the agent actually did. Using credentials it shouldn’t have had, finding an unknown flaw, and quieting the alerts meant to catch it is a textbook intrusion. It’s the same methodology a traditional red team follows, except no human was directing each step. It also shows the need to reassess the incentive structure for autonomous agents and their win-at-all-costs approach. When completing the task is all that matters, deception becomes a strategy, and that’s what we’re seeing as OpenAI’s agents game the very sandboxes built to contain them. All of this should change how organizations think about the AI agents they’re deploying internally. Most companies treat agents as productivity tools and give them broad access to code, cloud environments, and credentials, but very few have tested what happens when an agent behaves like an adversary. A sandbox is only a control if someone has tried to break out of it. Security leaders should bring agents into their offensive testing programs now. That means mapping what each agent can reach, red teaming the containment, and validating that monitoring still works when the thing being monitored has reason to avoid it. Whatever the courts decide, autonomous agents are now part of the attack surface, and they need to be tested like one.”

Matt “Mac” McKechnie, CTO and Co-founder, Corticle

“OpenAI deserves credit for disclosing this incident and publishing its research, because the whole industry learns from it. What gets lost in the public nuisance debate is that the Hugging Face hack was a security failure: an agent got hold of credentials it could use and found a way out of a sandbox that was supposed to be isolated.


We run autonomous agents in our own platform, and our rules are simple. The agents get no path to work outside the platform and very little creative license, and anything they aren’t fully confident about goes to a human. That’s basic security discipline applied to AI, and an agent that can suppress the alerts meant to catch it needs that discipline more than anything else in the environment.


Courts and regulators are going to keep pushing AI companies to prove their safeguards actually work, and the companies that built guardrails in from day one will have the easiest time doing it. For security teams, the practical step starts now: inventory every AI agent running in your environment and know exactly what credentials and systems each one can reach. But fundamentally, understand the architecture of your AI environment and assume breach within it, so you can map the paths an AI could take to break out and prevent it before it happens.”

California Attorney General Rob Bonta has sued OpenAI in San Francisco Superior Court. The suit alleges the company knowingly developed advanced AI models without adequate safeguards and calls it a public nuisance under state law. Bonta is seeking court-ordered changes to OpenAI’s practices and the surrender of related profits. The suit stems from a July incident in which an OpenAI agent escaped a sandboxed testing environment and accessed Hugging Face’s servers using stolen credentials and a previously unknown vulnerability. It also cites later findings that internal agents had suppressed monitoring alerts. OpenAI says it has published technical and independent reports and strengthened its safeguards since the incident.

I for one love the fact that OpenAI is being sued. It may force them to be more responsible because right now they are anything but. Also, I will say this one more time. If your defenses do not include accounting for AI, they need to.

Recent Critical Infrastructure Outages in US Commented On By Cloud Range

Posted in Commentary with tags on November 17, 2026 by itnerd

As of late, there have been a number of outages the the US and globally. The FAA and Microsoft come to mind as big outages. Bri Frost, Director of Product Management, Cloud Range has some thoughts on this phenomena:

“From Microsoft and air travel to banking and the attack affecting 178 organizations, people are seeing major disruptions unfold day after day. It’s understandable that they’re asking whether these events are connected and whether the systems they rely on are secure. When critical systems go down, the first public statement almost always says it wasn’t a cyberattack. These statements usually come out before the investigation is done. They may turn out to be right, but announcing a conclusion that early costs trust, especially when several major disruptions land in the same week. People notice the pattern, and ‘not cyber’ with no further detail doesn’t answer their real question. That question is whether the systems we depend on can hold up and whether the people running them are ready when they don’t. Organizations would build more trust by saying what they know, what they’re still investigating, and when they’ll share more, then following through. The deeper issue is readiness. Critical infrastructure isn’t going anywhere. Security teams have to be ready to respond to attacks on any and all systems, even outdated ones. Whatever the root cause turns out to be, outage or attack, the teams should be able to identify, respond and know how to restore systems quickly to reduce impact to their customers. You can’t build that muscle during the incident itself.”

Another Day…. Another Attack…. This Time With Optimizely

Posted in Commentary with tags on November 17, 2026 by itnerd

Let me introduce you to Optimizely. They bill themselves as “The AI platform to create and optimize every digital experience. So the experience-makers can do the work they love.” And they have just been pwned by a supply chain attack:

A supply chain attack targeting Optimizely has led to malicious JavaScript being loaded on numerous popular websites, including Zoom, CNN, Fox News, BBC, Salesforce, and GitLab, exposing visitors to scripts designed to steal cryptocurrency and payment information and exploit vulnerabilities in iOS devices.

Optimizely is a digital experience and experimentation platform used by organizations to test and personalize their websites, with its JavaScript files loaded into customers’ webpages to deliver this functionality.

The company confirmed to BleepingComputer that attackers exploited a vulnerability in its Experimentation account signup flow to inject malicious JavaScript into customer projects, causing websites using the platform to load malware.

I have commentary from Tyler Reguly, Associate Director of Security R&D of Fortra:

Years ago, websites were simple… a few lines of HTML, a markup language, that rendered information in your browser for human consumption. Today, websites might as well be complete software packages and, like software packages, they have dependencies. These dependencies come in the form of various JavaScript pages that are included from vendors and open-source projects designed to simplify life for the website developer, just as libraries simplify the lives of software developers.

The problem is that these dependencies are typically hosted with other vendors, which makes them part of your supply chain and opens the door for multiple breach vectors. Even if your environment is completely secure, every new dependency that you add is another window that needs to be locked, and you have no control over the lock.

That’s what happened here… Optimizely was breached, allowing attackers to compromise multiple other websites that may have been otherwise secure. If you visited one of these websites and entered credit card information during the hours of the breach, contact your provider immediately and cancel the card. If you visited one of the impacted websites from a mobile device, restart the device, I think that’s always sound advice in these situations.

At the end of the day, the JavaScript runs in the browser and none of the reports available indicate any browser escapes (other than the attacks on iOS), so the biggest risks were real-time, while the malicious JavaScript was available and, for the most part, financial in nature. With the breach contained, the malicious code removed, and websites operating normally, users should feel that they can once again visit these websites with confidence.

If you ever want to understand how intertwined websites are, the Computer Incident Response Center Luxembourg (circl.lu) has a great tool called Lookyloo that lets you see these interactions. I took one host off the list, a streaming service that I use, crave.ca and analyzed it with Lookyloo. While I visited a single URL with a single hostname, the analysis reveals that 42 hostnames were involved in loading that page with

Remember, you are only as secure as the people who you work with. So you have to make sure that they are secure as well. Otherwise, you get pwned as the case was here.

UPDATE: Justin Beals, CEO & Founder, Strike Graph, an AI-native GRC and compliance automation platform:

“Report URI and Cynet moved fast. They caught the injected code, named all 178 affected organizations and gave site owners exact strings to search for within a day. Optimizely’s notice confirms a vulnerability and leaves out everything customers need: how many projects changed, when the fix went in, and whether the attacker still has access. Smartsheet’s snippet got cleaned and then re-infected, so those questions matter.

Report URI had to crawl 1,611 Optimizely snippets to figure out who was affected. That tells you most of these companies had no current record of what third-party code runs on their own pages. An SBOM is supposed to answer that question for the software you ship, but this attack happened in code nobody builds or ships. The snippet updates itself every time a setting changes, and the attacker rewrote the payload several times in one day, so a document generated once and filed away would have shown a clean list the whole time.

Organizations need to apply the SBOM discipline to everything their pages load at runtime. Keep a live inventory of every third-party script on every page, record what each one should contain, and check it continuously against what actually gets served. A company with that inventory could have answered three questions in minutes on Monday: do we load Optimizely, on which pages, and did the content change. Most spent the day waiting on the vendor. Expect assessors and customers to start asking for that runtime inventory as evidence, especially on any page that takes card data.”

UPDATE #2: More commentary has come in:

John Strand, Owner, Black Hills Information Security

“There are a couple of things about this that are absolutely terrifying. The first is that we still don’t know exactly what the initial attack vector was. It’s been quite a while, and I’m increasingly concerned that they may not actually know how the attackers got in.

“The second issue is much bigger, and it’s something every organization needs to understand. Security is no longer just about one organization protecting its own assets. We are all part of a massive, interconnected ecosystem of systems, vendors, cloud platforms, and SaaS services. A compromise in one organization can have serious consequences for dozens, hundreds, or even thousands of others.

“And quite frankly, I don’t think most organizations have their heads around what their total attack surface actually looks like when you start incorporating all the third-party SaaS products and services they depend on. You can do everything right within your own environment and still find yourself exposed through a service you don’t directly control. That’s a very uncomfortable reality that the security industry needs to start taking much more seriously.”

ㅤ

BB King, Director of Pentesting, Black Hills Information Security

“A standards-based defense against this kind of supply-chain attack is Subresource Integrity (SRI). SRI allows browsers to detect when a script has been modified and then refuse to execute that script. It is designed for exactly this scenario.”

ㅤ

FusionAuth Appoints Phil Allen as Chief Revenue Officer and Eve Maler as Strategic Advisor

Posted in Commentary with tags on November 17, 2026 by itnerd

FusionAuth today announced Phil Allen as chief revenue officer and Eve Maler as strategic advisor. Allen will lead FusionAuth’s global revenue organization, overseeing partnerships, enterprise and commercial sales, sales engineering, and customer success. Maler will advise the company on product direction and identity standards.

The appointments come as FusionAuth accelerates its  international expansion. The company’s European annual recurring revenue (ARR) grew 72% over the past two years and now accounts for more than 30% of company ARR. In October, the company established its first European sales team in the U.K. to support its continued growth in the region.

AI adoption is driving that growth, as companies deploy agents faster than their existing identity systems can adapt. According to  FusionAuth’s 2026 State of AI and Identity report, 88% of organizations surveyed said AI deployment is outpacing their identity and security infrastructure.

Allen, who is based in London, has more than 25 years of experience in digital identity, cybersecurity, and fraud prevention. He joins FusionAuth from IDnow, the European identity verification company, where he was chief commercial officer, and was previously chief revenue officer of SecZetta through its acquisition by SailPoint. At Ping Identity, he led EMEA for nearly six years, through the company’s 2016 acquisition by Vista Equity Partners and its 2019 IPO. Earlier, he held sales leadership roles at Transmit Security, One Identity, CA Technologies, and BMC Software.

Maler is the founder of Venn Factory, an identity strategy advisory firm, and the author of Mastering Digital Identity: From Risk to Revenue. As CTO of ForgeRock, she helped guide the company through its IPO and eventual $2.3 billion acquisition by Thoma Bravo. Maler co-founded several identity standards, including Security Assertion Markup Language (SAML) and, recently, UMA for Agents.

Recast Introduces MCP Server to Bring AI Troubleshooting to Right Click Tools

Posted in Commentary with tags on November 17, 2026 by itnerd

Recast today introduced its MCP Server for Right Click Tools, bringing AI capabilities to daily IT workflows. Unveiled at Microsoft Ignite, MCP Server enables administrators to use natural-language prompts to investigate and resolve device and application issues. It creates a more complete picture of an organization’s entire fleet of devices by building on existing environment data with information from Right Click Tools. As AI becomes central to IT operations, these capabilities help IT teams address issues more efficiently and keep users productive amid rapid technological change.

For Recast customers, MCP Server can combine data from multiple Right Click Tools workflows to reduce manual research and provide a more complete view of device health. It honors existing permissions, routing, and licensing, helping organizations adopt AI while maintaining established access controls. This ability helps administrators make better troubleshooting decisions with real-time endpoint data. Unlike traditional endpoint management approaches that may rely on previously collected information, Recast’s MCP Server uses Right Click Tools to retrieve live data directly from devices, giving AI assistants current context to help administrators diagnose issues. Teams can use this approach to evaluate potential next steps based on conditions in their own environment rather than general guidance alone.

Recast’s MCP Server uses Model Context Protocol (MCP), an open standard for connecting AI assistants with tools and data, to extend the AI platforms organizations already use into Right Click Tools. With this connection, an AI assistant retrieves live environment data through Right Click Tools to answer an administrator’s question. This gives clearer insight into underlying issues, helping them identify root causes more efficiently.

Visa Ranked Most Trusted Among 24 Brands in Canada to Handle Agentic Transactions, New Research Finds

Posted in Commentary with tags on November 17, 2026 by itnerd

As artificial intelligence becomes part of everyday life for Canadians, trust will be critical to unlocking its next evolution: enabling AI agents to make purchases on consumers’ behalf. New Canadian findings from the Visa Trust Index reveal that 62% of Canadians surveyed would trust Visa to handle a payment made by an AI assistant, ranking it first among the 24 brands evaluated in Canada.

Meanwhile, Visa Canada’s Agentic Commerce Readiness Report highlights the gap between Canadians’ willingness to use AI for shopping and trusting it to take on purchasing tasks. The new report finds that nearly two-thirds of Canadians surveyed use at least one generative AI platform. Among current users, 83% would allow AI to assist with comparing products, prices and reviews. Yet, just 2% have used it to prepare or complete an order.

Canadians want to stay in control

Canadians see potential for AI to take on more of the shopping journey, but they want clear controls before giving it greater purchasing responsibility. Among current AI users surveyed for Visa Canada’s Agentic Commerce Readiness Report, 39% want the ability to pause or immediately turn off purchasing, 38% want control over which personal information AI can access and 35% want spending limits. Four in ten (39%) say they would always want to approve a purchase first when using AI.

Protection if something goes wrong matters, too. Among current AI users, 39% say they would feel more confident if they were not held responsible for unauthorized or fraudulent purchases.

Canadians remain cautious about AI making financial decisions

The Visa Trust Index further highlights Canadians’ caution around giving AI greater responsibility. While 40% of Canadians surveyed would trust generative AI to draft a difficult text message and 37% to manage their daily calendar, only 26% would trust it to buy items they regularly purchase. Trust falls further for more complex financial activities, including choosing stocks (17%) and planning for retirement (15%).

What this means for Canadian businesses

For Canadian retailers, Visa Canada’s Agentic Commerce Readiness Report points to a practical opportunity: helping shoppers find better value. Among current AI users, 42% identify finding lower prices or better deals as a benefit they already experience or hope to gain from AI shopping agents, making it the most-cited benefit. As AI takes on greater responsibility for purchases, clear approvals, spending limits and trusted payment experiences will be essential to building consumer confidence.


Research Methodology

About the Agentic Commerce Readiness Report

The Agentic Commerce Readiness Report is based on a survey that explores AI use and knowledge, concerns and motivators to use AI shopping agents. The survey of n=1,502 Canadians aged 18+ (including n=926 current AI users) was conducted by Angus Reid on behalf of Visa from September 9 to 14, 2026.

About the Visa Trust Index for Agentic Commerce

The survey was conducted by The Harris Poll through its Harris Poll Omnibus survey on behalf of Visa. In Canada, n=1,020 adults were surveyed in English and French from July 7 to 9, 2026, with respondents matched to the general adult population based on Census data. Between 503 and 519 Canadian respondents rated each payment brand on trust.

Fortra Launches Purpose-Built AI module for GoAnywhere MFT and Globalscape EFT

Posted in Commentary with tags on November 17, 2026 by itnerd

Fortra today announced the launch of its purpose-built AI module for GoAnywhere MFT and Globalscape EFT, designed to help organizations manage increasingly complex managed file transfer (MFT) environments with greater speed, consistency, and control. Currently available to select customers for testing and feedback, the AI module helps administrators accelerate setup, troubleshooting, documentation, reporting, and workflow analysis. General availability is expected in early 2027.

As organizations scale their MFT environments, operational complexity increases quickly. Workflows multiply, environments become fragmented, and administrators must analyze large volumes of logs, configurations, and execution histories. Tasks that traditionally require hours or days of manual review can now be accelerated through guided AI assistance helping teams identify issues faster, apply best practices more consistently, and make informed decisions with greater confidence.

The AI module is purpose-built for MFT workflows designed to support MFT experts. It expands access to best practices, accelerates analysis, and empowers administrators to stay in control through guided, auditable assistance with human oversight and control remaining in place.

For GoAnywhere MFT, the AI module introduces comprehensive capabilities for project creation and modification using natural language, guided diagnostics, and auto-generated documentation. Administrators can use natural language to create and modify projects before deployment, analyze workflows and logs to identify root causes and generate plain-language summaries that support audit readiness, knowledge transfer, and faster onboarding.

For Globalscape EFT, the AI module delivers AI-assisted capabilities that offer critical event diagnostics, event rule analysis, and dynamic reporting. Administrators can drill into event findings, analyze rules and logs, and generate summaries of trends and risk signals for operations and compliance review.

Together, these capabilities give both new and existing customers a practical path to faster, more consistent MFT operations. Outputs are designed to be reviewable and auditable, with administrators remaining in control before changes are applied. Capabilities such as human review before deployment, role-based access, identity enforcement, network controls, and rollback-ready workflows help ensure AI-assisted recommendations support expert decision-making without bypassing established governance processes.

The technology supporting the AI module has also enabled Fortra to develop migration capabilities that help organizations transition from competitive MFT platforms to GoAnywhere MFT or Globalscape EFT. Planned for release later in 2027, these capabilities can accelerate the analysis and conversion of existing workflows, reducing migrations that traditionally take months to a matter of weeks. It minimizes the time, effort, and disruption associated with replacing a legacy or competing MFT solution, allowing organizations to modernize without placing the full migration burden on already constrained IT teams.

To learn more about Fortra’s AI-enabled MFT capabilities visit GoAnywhere.com and Globalscape.com.

Finite State Appoints Mukesh Singh as Chief Revenue Officer

Posted in Commentary with tags on November 17, 2026 by itnerd

Finite State today announced the appointment of Mukesh Singh as Chief Revenue Officer. Singh has more than 30 years of experience building global sales organizations across cybersecurity, AI, and enterprise software, most recently as Chief Revenue Officer at Deep Instinct.

Singh will lead Finite State’s global revenue organization, including sales, partnerships, and customer expansion, as the company scales. His appointment comes on the heels of Finite State’s $27 million Series C funding round and its recent win in IoT Evolution World’s 2026 IoT Security Excellence Awards, underscoring the company’s growing momentum in the product security market.

Singh joined Deep Instinct in 2024 and became Chief Revenue Officer in 2025. He previously held senior sales leadership roles at TrueFort, ColorTokens, Confluera, Illumio, Virtual Instruments, and IBM. Singh has built and led revenue organizations at 14 technology startups, taking teams from early traction to scaled, repeatable growth. He began his career as an engineer at Hitachi Data Systems and EMC and studied electrical engineering at Ryerson Polytechnical Institute, now Toronto Metropolitan University.

For companies that build connected products, from medical devices and connected vehicles to industrial control systems, product security has become a core part of bringing products to market. Regulations such as the EU Cyber Resilience Act, FDA premarket cybersecurity requirements, ISO/SAE 21434, and IEC 62443 now ask manufacturers to demonstrate the security of their products before launch, and customers increasingly expect the same in their purchasing decisions. As requirements tighten worldwide, manufacturers that make security a priority will be best positioned to compete in key global markets.

These pressures have elevated product security to a board-level concern, and manufacturers are accelerating their investments. The IoT security market is projected to reach $80.3 billion by 20311, and Finite State sees growing demand in every sector it serves, including medical devices, automotive, industrial and OT, energy and utilities, consumer IoT, telecom, and defense.

Singh’s appointment follows the addition of Ann Miller as Vice President of Marketing earlier this year, further strengthening Finite State’s leadership team as the company scales its global operations.

88% of top AI companies lack public vulnerability reporting programs

Posted in Commentary with tags on November 17, 2026 by itnerd

While OpenAI is increasing its bug bounty rewards to as much as $172,000, most major AI companies still lack a publicly identifiable way for security researchers to report vulnerabilities, according to the Cybernews AI Trustworthiness Ranking.

The Ranking, which evaluates 500 leading AI companies, found that only 12% (62 of the analyzed companies) have a public bug bounty or coordinated vulnerability disclosure (CVD) program.

Key findings:

  • Among the companies with public programs were OpenAI, Google, Microsoft, Meta, Anthropic, Mistral AI, Perplexity AI, and Hugging Face.
  • 438 companies had no publicly identifiable program, among which were Midjourney, ElevenLabs, DeepSeek, Cursor, Runway, and Suno.

With users increasingly sharing sensitive information, including health details and work documents, with AI tools, undiscovered security vulnerabilities could put that data at risk.

For more information on the AI Trustworthiness Ranking, click here.