Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Your Monday catch-up, with a security lens. This week: Microsoft makes its AI agents generally available at Build, and the ShinyHunters Canvas breach escalates into what's being called the largest education-sector breach on record. On security: certificate authority DigiCert is socially engineered into issuing fraudulent code-signing certs, and Zara and an NVIDIA cloud-gaming partner are breached. Confident your staff can't be talked into handing over access? Visit www.kinsoft.com.au to talk through your security and IT needs. Sources: Microsoft; The Hacker News; CNN; Comparitech; CM-Alliance.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Imagine building like the ultimate completely impenetrable fortress. I am talking titanium walls, laser grids, biometric scanners that check your retina and your fingerprints. The whole nine yards. Right. The whole nine yards. You spend billions on the architecture and it is mathematically perfect. And then on day one, the security guard stationed at the front desk gets a phone call. from someone pretending to be, I don't know, the fire inspector. And the guard just casually props the titanium backdoor
open with a brick. Wow. Yeah, it completely invalidates the billion -dollar laser grid, right? Because the technology worked flawlessly, but the overall system still suffered a, well, a catastrophic failure just because of that human element. Welcome to the Deep Dive. Today, we are looking at a recent, incredibly revealing snapshot from the tech landscape. It's the Tech Talks with Kinsoft update from May 11th, 2026. A really fascinating brief. It is, because this single document captures
this terrifying paradox about where we are right now. You know, we are building these mathematically perfect digital fortresses, but we are just leaving those titanium doors. propped open everywhere. Yeah, that's exactly it. On one hand, our technology is becoming unbelievably autonomous and capable. But then on the other hand, our biggest vulnerabilities are still fundamentally human. It really is kind of the defining tension of this era we're in. When we analyze this Kinsoft source material,
two major interconnected themes emerge for you to really think about. Right. First, we have the official mainstream arrival of autonomous AI agents. And then running totally parallel to that massive leap forward, we have a series of just catastrophic cyber breaches. Huge ones. Yeah. And the narrative thread tying this all together is that as we hand over more power to technology to act on our behalf, you know, the human gatekeepers managing these systems are increasingly becoming our absolute weakest link. Okay, let's unpack this because we have to start
with a big announcement from Microsoft's recent build conference. Oh, absolutely. According to the Kinsoft update, Microsoft has officially made the, quote, agent era a reality. They announced that their Agent 365 platform, along with this entirely new agent framework and a whole stack of in -house AI models, has moved to general availability. Right. General availability is the key term there. Yeah. Plus they rolling out a new unified intelligence layer across all the developer tools. So we are officially out of
the experimental sandbox here. But, you know, what does that actually mean for you, the listener? We hear words like framework and layer. all the time in tech we do but if we connect this to the bigger picture the shift to general availability is a massive milestone this isn't uh beta testing for a few select tech enthusiasts in a basement anymore right this is a fully supported enterprise -grade product rolling out to the global mainstream and the transition here the really important part is moving from conversational ai to agentic ai okay break that down a bit well For the last
few years, we have basically been talking to incredibly smart chatbots, right? You ask a question, it generates some text. But an AI agent is an entity designed to take real independent actions inside your actual software ecosystem. So rather than just like generating a draft of an email that I have to manually copy and paste, the agent actually reads my inbox, decides an email needs to be sent, writes it, and literally hits the send button for me. that is the practical application yes it's doing the work yeah and to understand
how that actually functions we have to look at that unified intelligence layer you mentioned earlier okay think of previous ai as like a brain trapped in a jar a brain in a drawer right it was really smart but it couldn't touch anything yeah this new unified intelligence layer acts like a nervous system connecting all your different applications it provides the digital roads for the agent to travel between your word processor your email client your corporate database. And the framework itself gives the AI the hands to actually push buttons and move files across those
different software environments. It is honestly like hiring a brand new, incredibly fast, hyper -efficient digital employee. You bring them on board and suddenly you have this massive workforce multiplier. Oh, massive. But wait, I have to push back here for a second. Think about it like corporate credit cards. Okay. If you give a corporate card to a human employee, you have an HR file. You've done a background check. You have a manager actively watching their spending. Hopefully, yeah. Hopefully. But if we are handing out digital ID badges to these AI agents, you know, giving
them access to our sensitive corporate environments, how do we actually govern a machine's identity and its access rights? Oh, the million dollar question. Right. Because imagine handing out 10 ,000 corporate cards in a single second to invisible digital workers. Yeah. How do you even begin to track who's doing what? You've hit on the exact vulnerability the Kinsoft update explicitly warns about. Every single agent you deploy is a new non -human identity with access rights that absolutely must be governed. It's just a
massive headache waiting to happen. It is. When an AI agent logs into a database to pull records, the system needs to know if that specific agent is authorized. If an attacker manages to hijack an agent or, say, trick an agent through a malicious prompt. Oh, wow. The attacker inherits all the machine's permissions. We are rapidly expanding our attack surface by creating thousands of these non -human identities that basically hold the keys to our digital kingdoms. Which means the platforms managing all these identities and all
this access need to be incredibly. secure, like flawlessly secure. And that introduces a major structural problem. Right. Because giving AI new identities and deep platform access requires those underlying platforms to be flawless. But the reality of our current infrastructure is, well, it's anything but flawless. Which brings us to the architecture of the cloud itself. That feels like the perfect bridge to the next massive story in the Kinsoft update. Since we're giving these agents the keys to our cloud platforms, we really need to look at how fragile those platforms
actually are right now. They're more fragile than people think. Yeah, and the nightmare scenario unfolded recently at Instructure. For those who don't know, Instructure runs Canvas, which is this massive learning management system used by countless schools and universities. Almost everyone in education uses it. Right. Well, the threat actor, a crew known as Shiny Hunters, managed to breach the system. And they didn't just quietly steal data. No, it was very loud. They actively defaced the Canvas login portals of roughly 330 different institutions. They stole
data belonging to hundreds of millions of users. And the situation was so dire that a ransom deal was eventually struck. It's just staggering. This incident is being called the largest education sector breach on record. Unbelievable. And to really understand the sheer scale of the devastation here, we have to examine how these services are actually delivered. Canvas operates as a massive SaaS platform, software as a service. Right. SaaS. The central lesson here from Kinsoft is the terrifying scale of vendor breaches. Break
down that scale for us a bit. Like, why is a SaaS breach fundamentally different from, say, a traditional network hack from 10 years ago? Well, in the past, if a university wanted a digital portal for its students, they bought their own physical servers, they installed them in a dusty basement on campus, and they ran the software locally. Okay. An attacker would have to breach University A's firewalls and then start completely from scratch to separately breach University B's firewalls. It was a lot of manual work for the hacker. Exactly. But with a multi -tenant
sauce model, all those universities outsource their infrastructure to a single centralized vendor. Instructure is the vendor. It is essentially a giant apartment building where everyone shares the exact same foundation and the same plumbing. So the attacker doesn't need to break into 330 separate university servers. They just need to find like one vulnerability in the master key system of the apartment building. You got it. And once that single central hub is compromised, the disaster cascades instantly to every single tenant. The economies of scale that make sauce
so affordable and so efficient. Which is why everyone uses it. Right. Those same economies of scale also create massive single points of failure. It's kind of terrifying when you think about it. You drop one massive cloud -shaped basket. And 330 institutions lose hundreds of millions of user records. I mean, grades, personal information, financial data, all at the exact same time. And think about the ransom negotiation. Oh, God. How does a single vendor negotiate a ransom when the data belongs to hundreds of independent universities? The legal and logistical nightmare
is just staggering. Seriously. So what does this all mean for the listener? I mean, whether you are a student, a professional, or running a business, we all rely on massive... platforms every single day. If the platforms holding our most sensitive data can be breached and held for ransom by crews like Shiny Hunters, is any data actually safe? Like, is this just the cost of doing business in 2026? It forces a very sobering reality check about vendor risk. You can have the most draconian,
hyper -secure password policies in your own organization. But if your vendor's security fails, your data is completely exposed. Wow. It completely shifts the burden of trust away from your internal IT team and straight on to a third party. But let's play devil's advocate for a second. Let's say a university gets spooked by the campus breach and decides to pull all their data out of the cloud. They move everything back to those isolated servers in the campus basement. They are completely safe now, right? They are not. Not at all. No, because even offline, isolated servers still need to download software updates, right? To
patch known bugs. And that exposes a completely different... much deeper vulnerability. Oh boy. Breaking into a software platform like Canvas is one thing, but breaking the system that tells the internet what software to trust in the first place, that is a foundational crisis. Which perfectly leads us to a story that might seem, I don't know, a little abstract at first glance, but is arguably the most unsettling part of this entire Kinsoft update. Definitely. We're talking about the breach at DigiCert. According to the source, attackers successfully manipulated DigiCert
into issuing fraudulent code signing certificates. And then the attackers used those fake certificates to digitally sign their malware, making the malicious software look entirely legitimate to literally anyone who downloaded it. What's fascinating here is the underlying mechanism of Internet security that this attack actually exploited. We really need to talk about the concept of a trust anchor. A trust anchor. Yeah. Whenever your computer downloads a piece of software or, say, an app, your operating system does a quick background check before allowing it to install.
Right. It asks, who created this file and has the code been tampered with since it left the developer's computer? It's looking for that little pop -up that says, like, verified publisher. Exactly. And the way it verifies that publisher is through public key cryptography. Certificate authorities, like DigiCert, are the organizations trusted by Apple, Microsoft, Google to issue those cryptographic certificates. Okay, so they're the authorities. They are the digital passport agencies of the internet. When DigiCert signs a certificate for a software developer, it is
mathematically vouching for them. Gotcha. The operating system sees the DigiCert signature and applies a fundamental hard -coded rule. This file is digitally signed by a trusted anchor. Therefore, it is safe to run. So it's like a master counterfeiter. Right. But instead of just printing fake money in a basement and hoping a cashier doesn't look too closely. Right. The counterfeiter actually manages to trick the official government mint into stamping their fake money with a real mathematically perfect watermark. The money itself is fundamentally counterfeit,
but the security features. are 100 % genuine. That captures the severity perfectly. By issuing those fraudulent certificates, DigiCert unknowingly gave malware a VIP pass right past the bouncers of every single computer security system. The fundamental guarantee of, is this digitally signed, was just completely broken. Wait, wait, you're losing me here. I have to push back on this. Sure. We are talking about DigiCert. A company whose entire existence is based on mathematically
perfect billion dollar cryptography. Yes. We're talking about prime numbers so large that it would take a supercomputer like a million years to crack them. And you're telling me the bad guys bypassed all that math just by calling customer service and asking nicely. I know. How does a help desk worker even have the authority to issue a cryptographic certificate? It sounds totally absurd, but it highlights a massive blind spot in how we structure security. Customer support portals have administrative overrides. Oh, of course they do. Right. They're designed to help
legitimate customers who lose their keys or need expedited service. The attackers didn't use some brilliant zero -day hacking code to break into DigiSearch cryptographic vaults. No code at all. No. The Kinsoft update explicitly states they used social engineering through the support portal. They just manipulated a human being who had authorized access to the tools that generate those certificates. So the foundational trust anchors of the entire Internet really were defeated by a smooth talker on the help desk line. They were. Honestly, it
makes perfect sense from the attacker's perspective. How so? We spend billions analyzing code, patching servers, and configuring firewalls. But attackers realize that the path of least resistance is almost always human psychology. Yeah, why break the math when you can break the person? Exactly. Why spend months trying to crack an encryption algorithm when you can just convince an employee that you are a frantic developer facing a massive deadline who needs a certificate reissued immediately? Wow. Social engineering relies on manipulation,
you know, creating a false sense of urgency, leveraging fake authority, or just exploiting natural human empathy to trick a person into bypassing all the rules. Which brings us to the ultimate takeaway from the Kinsoff brief, and the realization that the human element is sort of the universal bypass here. And this wasn't just an isolated incident at DigiCert. Not at all. The update heavily emphasizes recent breaches at the global fashion retailer Zara and at an NVIDIA cloud gaming partner. And the common thread, Kinzov calls social engineering, quote, The week's
quiet villain. A great phrase. Let's dive into how these specific breaches happened. Because Zara and NVIDIA are massive companies with huge security budgets. They have multi -factor authentication, identity protocols that works. How does a smooth talker bypass all of that? The tactics are incredibly sophisticated, yet entirely non -technical. In scenarios like the NVIDIA Cloud Gaming Hartner breach, attackers often utilize a technique called MFA fatigue combined with IT impersonation. MFA fatigue. Yeah. So they acquire a target's username
and password, often from an older, completely unrelated data leak. They try to log in, which triggers a multi -factor authentication request to the real employee's phone. The classic press one to approve this login notification. Yes. And the attacker will spam that notification to the employee's phone 50 times in the middle of the night. Oh, that's incredibly annoying. Right. The employee is annoyed, confused, and just wants their phone to stop buzzing at two in the morning. But they usually don't click approve. Good. But then the next morning, the
attacker calls the employee, spoofing the phone. number so it looks like the internal IT help desk oh no the attacker says hey we noticed some weird login activity on your account last night we are trying to secure it but I need you to approve the notification I'm sending to your phone right now to verify your identity oh wow so the employee thinks they're talking to the good guys right I think clicking approve is securing their account when in reality they're letting the attacker straight into the network exactly and similar tactics are used against Exploiting
that authority. Yes. They bypass the technical identity protocols by manipulating the human desire to be helpful or, honestly, just the human fear of getting in trouble with the boss. It is the brick. propping open the titanium door all over again. It really is. The laser grid works perfectly, but the security guard was tricked into turning it off. You know, if we synthesize everything we've discussed today, a very stark picture emerges. We are deploying incredibly advanced autonomous AI agents through platforms
like Microsoft's Agent 365, giving them their own digital corporate credit cards to roam our networks. We are centralizing hundreds of millions of user records in massive SaaS infrastructures like Canvas, creating incredibly lucrative targets. The technological stakes have never been higher. Never. Yet despite all of this advanced technology, attackers are succeeding through completely low -tech methods. Right. using convincing phone calls, manipulated emails, and support portal trickery to bypass the strongest digital defenses we can possibly build. The technology isn't failing.
The human gatekeepers are being manipulated into opening the doors. And the Kinsoft update ends with a very direct piece of advice that connects this right back to you, the listener. Their sign -off is stay patched, stay skeptical. Good advice. Because whether you were running a massive corporate IT department or you're just sitting at home trying to figure out if that text message from your bank is a phishing scam, the human element is the final line of defense. It absolutely is. Critical thinking is a required security control
now. Knowledge is most valuable when it is understood and applied. We can build the most robust AI frameworks and secure our sauce environments with mathematical precision. But if we don't train ourselves to view our digital interactions with a healthy dose of that recommended skepticism, the technical defenses simply do not matter. They're useless. Pretty much. The attackers will always target the easiest way in. And right now, that is us. So to quickly recap this incredibly dense snapshot of our tech landscape, we are officially entering the era of. autonomous AI
agents taking real actions inside our software nervous systems. At the same time, we are witnessing the terrifying single point of failure vulnerability of massive SaaS platforms with breaches affecting hundreds of millions of people at once. And underneath it all, the very cryptographic trust anchors of the internet are being bypassed, not with complex code, but with simple human conversation and psychological manipulation. It is a wild time to be navigating the digital world. It certainly is. The landscape requires a constant active
reevaluation of where our true vulnerabilities actually lie. And as we wrap up today's deep dive, I want to leave you with a final lingering thought to mull over. We spent a lot of time talking about how easy it is to socially engineer a human being like the support staff at DigiCert or the employees at an NVIDIA partner into making a terrible mistake. But think about those new autonomous AI agents Microsoft is rolling out. These agents have their own digital identities, their own access rights, and the ability to interact with a broader digital ecosystem on our behalf.
What happens when those AI agents become the targets of these exact same social engineering tricks? We know a smooth talker can manipulate a tired human help desk worker by creating a false sense of urgency. But as these AI models read our emails and interact with outside data, I mean, can you fast talk an AI into handing over the keys to the kingdom? That is a terrifying thought. That is something to think about the next time you authorize a digital agent to manage your inbox. Thank you for joining us on this deep dive. Stay skeptical out there, and we will
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Japanese home goods company Iris Ohyama (online store operator Iris Plaza) found a suspicious file in an external...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Australian online retailer Who Gives A Crap told customers on 7 November 2026 that their information was accessed...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Last week in tech, 7 to 14 November 2026. Microsoft's November Patch Tuesday (10 November) fixes well over a...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Chainalysis disclosed on 30 October 2026 that an external actor using "offensive agentic techniques" accessed...
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! NSW Health reported unauthorised access to its Oral Health Fee For Service Scheme website on 29 October 2026;...