Transcript 0:00 [upbeat music] Welcome back to another edition of Vanguard Radio. I'm your host, J. Richard Jones. 0:13 Today we're gonna be talking, uh, about AI and its role, as well as security, the challenges, and, uh, some of the opportunities and solutions as well. 0:22 Our guest that's joining us is with a company that is leading AI, and when I asked him before the show about what does he see some of the biggest challenges being, it's really there's a revolution going on with AI, and a lot of the adversaries are becoming a lot more sophisticated in what they can do. 0:40 But the good news is we're witnessing a flipping in the legacy paradigm, and tools and automation are really helping people to become a lot more proactive. 0:50 The company is Palo Alto Networks, and it is the world's cybersecurity leader. Our guest is Mr. Daniel Cruz, and he is Vice President of Public Policy and Government Affairs with Palo Alto Networks. 1:03 His responsibilities do include Canada, and he thinks a lot about what's going on within the Canadian marketplace, the Five Eyes, and how everybody can work together. Daniel, welcome to the show. Thanks, J. 1:15 Great to be with you today. You know, AI is really top of mind in so many ways, and, and certainly on the security side and policies and procedures. It's top of mind within industry. It's top of mind in... 1:27 with the population in general in how it's being used. How can organizations enable the safe use of AI, or do you think they can? Absolutely. 1:37 Responsibly deploying AI is gonna be transformational for a range of industries. Uh, and I'm a little bit biased of course, but I think cybersecurity is top of the list. 1:46 I, I think the reason for that is cybersecurity has historically really been a data problem. 1:51 You have all these different tools and sensors that give you different telemetry of what is happening across your digital infrastructure, but then the bottleneck has been at that ingestion point, that analyst who's overworked and burnt out and is trying to synthesize and distill and normalize all those different data streams, make sense of them in real time, and then act upon that to actually fortify their networks. 2:13 Doing that through manual processes is like running uphill on a treadmill. You're just never gonna get ahead. The adversaries are always upper hand. That has been the legacy frustration of the cybersecurity industry. 2:24 Flipping that legacy paradigm and giving defenders the upper hand with AI-powered automated cyber tools, like that's really the moment we face right now, and so that's what's really exciting about my job as we work with, you know, allied governments around the world. 2:38 We work with large enterprise customers, critical infrastructure. 2:41 All the legacy frustrations of drowning in data, but not actually making their networks any safer, we do live at a moment where we can flip that paradigm, and we're seeing organizations to do that right now. 2:52 And the results aren't really hypothetical. They're real. The rubber's already hitting the road. Uh, and it's proving to be transformative. So that's really creating empowerment and efficiencies. Totally. 3:02 And if I can give you a stat here just to really highlight what I'm talking about. 3:05 You know, we, we often try to distill cybersecurity, and for your, your listeners here, you know, into metrics that your average CEO or board member can understand, right? 3:13 Your average CEO or board member does not need to be a technical or cybersecurity expert, but they need to know the right questions to ask their experts so they can have a good sense of the overall cybersecurity health of their enterprise and their organization. 3:25 A- and we really think the two questions they should be asking are, "What is my organization's mean time to detect, and what is my organization's mean time to respond?" 3:34 Those are two terms that describe exactly what they sound like. 3:37 How long does it take you to detect that something might be wrong in your systems, and how quickly can you respond and resolve the thing that might be wrong? 3:45 Historically, the mean time to detect for organizations was oftentimes in many days, and the response times were often in weeks to months. 3:53 So we built an AI-powered tool that stitches and synthesizes all of your different data streams together, separates the signal from the noise, automates a lot of response through automated playbooks, and puts the defenders back in charge. 4:05 Before that tool was installed in our customer base, the average mean time to respond was two to three days. After, 14 minutes to two hours. So think about that for a second. That's not hypothetical. 4:16 That's the real world example of when you're living in the legacy paradigm, drowning in data, but can't do anything with it. 4:22 You're taking two to three days to respond to an incident, when adversaries are oftentimes only taking five to six hours to get your crown jewels, your data, and get out of there. 4:30 But after installing the best AI-powered systems to, to actually give you the upper hand, to actually make sense of that security data, that has now been compressed to 14 minutes to two hours. 4:39 So it's not an exaggeration to say that this is truly flipping the paradigm. It's turbocharging your response. It's putting you from being behind to being ahead. So that's what we're seeing right now. 4:48 Uh, and if my excitement and enthusiasm is, is coming through at all here, it's because these are transformative outcomes. We don't need to squint our eyes and, and wait for something to happen in a decade. 4:59 It's happening now. The rubber's hitting the road a- and we're seeing transformative outcomes across the public and private sector alike. 5:04 So there is, despite all the, the headlines you see, there still is an optimistic message in cybersecurity, and AI is a big part of that. So some of the stats you just quoted, that's great. 5:15 Does that kind of fall under the banner of secure by design? I mean, we hear that talked a, a lot about when it comes to cybersecurity these days. Yeah, absolutely. 5:24 And secure by design is, uh, is a principle that's been around for a while. 5:27 Perhaps one of the most publicly visible discussions about secure by design is something that the Cybersecurity and Infrastructure Security Agency, CISA, within the US Department of Homeland Security, has been leading. 5:39 I am a, uh, a proud alum of that agency, so I appreciate the good work that they're doing there. But CISA has this voluntary secure by design pledge that has seven different pillars underneath it. 5:49 Uh, I think as we sit here today, over two hundred and fifty organizations have actually signed that pledge. And so, you know, CISA is using its convening authority to... 5:58 And, and, and almost all those organizations are, are multinational organizations that certainly have a large presence in Canada. 6:04 And so CISA is using its convening authority to help drive better security on the front end. 6:09 I think, you know, central to the secure by design pledge is this concept that, you know, for too long, security was kind of bolted on after the fact. 6:16 People built the infrastructure, they deployed the digital tools, and then kind of had this oh, shoot moment of, well, how, how are we actually going to make the, the damn thing secure? 6:23 Uh, and secure by design very much tries to shift left as, as a lot of folks in the industry like to, to say, uh, and, and think about security from the outset. 6:32 So, so that is a big part of the secure by design push, and, and we, we are signers of that pledge, and, uh, and that's central to the work that we do. We also want to take things to the next level. 6:42 In, in addition to the secure by design effort, which CISA and other organizations are big proponents of, we've introduced this concept of secure AI by design, which in, in many ways could be seen as a, an extension of existing secure by design efforts, but really recognizing that in the age of AI, there's a little bit of a new lens through which we need to look at historical data challenges. 7:03 And, you know, at the end of the day, I, I view as data is the currency of the twenty-first century. 7:07 The confidentiality, integrity, and availability of data underpins really every critical infrastructure resilience conversation. 7:15 It underpins every personal privacy conversation about all the information you trust government agencies to protect on your behalf. And so it, it really is just all about the data at the end of the day. 7:24 Well, what is AI as a new ingredient? It's really a, it's an accelerator and turbocharger of just the velocity and importance of all that data. You know, the, the attack surface of this AI infrastructure. 7:35 You've now got all this training data, and you've got the models, and you've got the weights. 7:39 And so it's basically taking the, uh, e-existing ecosystem of, of where data is stored and transited and, and shaking up that snow globe and, and, and just, you know, accelerating a-all of that. 7:50 Uh, and so recognizing that, what is-- what, what do secure by design principles mean in the era, uh, of AI? 7:57 Because there's just this whole new digital attack surface of just all these new buckets of data that we need to secure. And so we've really boiled it down, secure AI by design, to really three overarching categories. 8:08 And the first is, you know, organizations need to have some consistent way to govern AI usage for employees to prevent data leakage. 8:16 Our, our stats show that somewhere between fifty and sixty percent, uh, of employees at organizations are tinkering with and experimenting with various gen, gen AI apps, which is great. 8:26 They're, they're doing so a-almost exclusively with good intentions. You know, they're trying to find applications and things to make them, them more efficient in, in, in their work. But what are those apps training on? 8:36 Could a well-intentioned effort to deploy some new app to maybe try to streamline a business process inadvertently be training on proprietary customer PII within your enterprise? 8:46 And so you need to have some way to govern the AI usage, so it's not just running amok, and so that there's confidence that the data is where it should be and is not where it should not be. 8:55 And then the second is we need to secure every step of the AI app development life cycle and, and supply chain. 9:00 As again, it's just the, the number of different vectors of the way that data is being transacted and handed off through all these different actors across the, the, the value chain. 9:10 It's just more complicated in the age of AI. So we need to make sure that we secure that, uh, in, in every step. 9:16 And then lastly, during what we call runtime, which is when an AI app is running, we need to then protect that AI data from unauthorized access and, and leakage at all times. 9:25 You know, you've heard about things like prompt injection attacks, which is if a malicious actor is able to insert something into either the training data or the prompt for some sort of gen AI application, then the output is going to be different than what the user anticipated. 9:39 But they're not aware that there was manipulation, uh, or poisoning of that prompt. 9:43 So those are really the three sort of elements, uh, of secure AI by design, governing AI usage for employees, securing every step of the AI app development life cycle and supply chain, and then protecting AI data during runtime. 9:56 So it's really an exciting time, both on the secure by design effort and then now sort of the next frontier, secure AI by design. 10:04 You know, as you're going through those points, uh, it sounds like there almost needs to be some sort of international standard for the policies, the procedures, and the testing, and I guess the output and what you do with that data. 10:17 Is there like, uh, an ISO standard that's already in place for this? Or, you know, is it like the, the Wild West and it's in the stages of development? 10:26 Uh, I wanna say we're, we're sort of mid, mid Wild Wild West, but good work has been done and, and there's some promising efforts underway. 10:33 Uh, and so, you know, NIST, which similar to ISO in, in, in some manners, ha-has some efforts underway. 10:40 They have some compendiums to their existing cybersecurity framework and to their existing secure software development framework for gen AI that includes some controls in there. So, so that's, that's one element. 10:52 Uh, there also is, uh, a really, and you know, I, you know, pr-perhaps, uh, a guidance that I think is most promising. 10:58 There is a, a Five Eyes guidance, and so it has the, you know, all, all five cyber centers in all Five Eyes countries as sort of the, the banners, the logos on, on the top of it, uh, around secure AI deployment, uh, that runs through, uh, a range of these concepts. 11:12 So generally aligned with what I just laid out across those three, three buckets. 11:16 And so, you know, that's still at this point voluntary guidance, but, you know, could that become something a little bit more rigid in terms of more of a NIST or ISO-like set of, you know, security controls? 11:25 Basically taking what they have in that guidance and maybe just double-clicking on the concepts and, and adding the technical controls underneath that. 11:31 So I do think we're at that moment where we're maybe mid Wild Wild West, but there's enough promising areas of activity, all with international cooperation, that, you know, I think the next year you're gonna see a lot happen. 11:42 Well, that's certainly a, a relief. 11:44 Daniel, you mentioned supply chain integrity, uh, as, as being an area of importance in, here in Canada, and, and it certainly is, and I would imagine it's a big issue all over the world. 11:54 Can we, uh, expand on that a little more? How is Palo Alto Networks working to ensure that there's integrity in that process? You know, it's interesting. 12:03 When I was at, when I was at CISA in, in the, the 2018 to 2021 timeframe, I increasingly found myself saying kind of on the, the national security circuit that cybersecurity risk management and supply chain risk management were becoming inexorably linked. 12:17 They sort of were becoming two sides of the same coin from my vantage point because I think what we have seen is that, you know, it's, it's no longer just about protecting an enterprise's network. 12:27 To protect their network, you actually have to protect the hundreds of digital things that constitute the network. 12:35 So it's not like Acme Organization just has some m- monolithic network that needs to be protected from a cyber standpoint. 12:41 It's actually the hundreds of either on-prem or SaaS-based applications and other tools that make up their digital estate. 12:49 And so understanding the complicated web of vendors that either physically or virtually provision your digital infrastructure is just central to the cybersecurity of your digital infrastructure. 13:01 So in that sense, uh, I think that cybersecurity risk management and supply chain risk management are, are really intertwined disciplines. They're, they're two sides of the, of the same coin, so to speak. 13:11 And so we take this as a company very seriously, and actually we really were, we're pleased that, that NIST did a, a full case study on our supply chain risk management practices. 13:19 I think a huge part of it at the get-go is, is just illumination and, and understanding of your supply chain. 13:25 So, you know, I sort of talked about when, when you think, you know, just on a daily basis, uh, as you go about your, your life and your profession and, and your job, the number of different apps and digital tools made by different vendors that you interact with on a daily basis just between 9:00 AM and 5:00 PM, it's dozens, if not 100. 13:42 But you so seamlessly sort of shift between those that it kind of feels like, "Well, yeah, that's just my laptop." 13:47 Well, that's your laptop that you're, you know, you're probably using 30 different applications, uh, across several dozen different vendors. And so illumination into that, uh, is, is really important. 13:57 Uh, and sort of what you ask of, of your suppliers is, is something that, you know, we take seriously because then, you know, large purchasing power can sort of drive behavior. 14:05 And, you know, we're, we're, we're seeing that, uh, in, in, in large entities, and I'll use a, a US example here, but there are other examples. 14:11 Like, you know, we have a, a new program that's going live in our Department of Defense, CMMC, the Cybersecurity Maturity Model Certification, which recognizing that the defense industrial base has 250,000 entities. 14:22 There's sort of a recognition that, well, if you're able to use the power of DOD's purchasing power to drive supply chain activity, you know, that can move the needle. 14:29 I know there are similar efforts in Canada, and Bill C-26 is something that we're tracking very closely and, and in supportive of what it's, it's trying to catalyze from that perspective. 14:37 So I think you're seeing a lot of... 14:38 that the recognition that cybersecurity and supply chain risk management are intertwined, that all entities, sort of large and small, need to have a dedicated effort to understand and to illuminate the supply chain. 14:49 And then within their sort of contractual operations, procurement is not just a check-the-box thing, it's actually an opportunity to drive practices. 14:57 All of those are things that we take really seriously, uh, and, and a reason that we were, you know, proud to be recognized by NIST for our supply chain risk management. 15:04 Yeah, you know, um, one of the areas I just wanna touch upon is, um, secure internet browsers. Um, what are the immediate benefits for users, and, and, and what, what do you see as the benefits from your perspective? 15:18 I think it closes, uh, a visibility and security control gap that had emerged. 15:23 There, there has been so much necessary investment in cybersecurity solutions, uh, across all different ranges, uh, of network infrastructure, from the, the network edge, to endpoint agents on devices, to cloud security tools that ensure that your cloud environments are appropriately configured. 15:41 But for a long time, the sort of black security hole was the browser. People then went into their browser and did all sorts of transactions. 15:49 I mean, think about the data and the sensitive data and, and the financial transactions and everything that is happening just within the browser. That is increasingly becoming one of the key sort of security bottlenecks. 16:00 And from a security perspective, there historically had been, uh, a, a lack of sort of visibility and, and enforcement of security policies within that environment. It was just sort of this cabin off to the side. 16:11 And so secure browsers actually allow you to integrate the security posture of the rest of your digital estate from, you know, what you have in the network and the cloud and, uh, and onwards into the browser, where increasingly users are spending, uh, a l- a lot of their time. 16:24 Uh, you know, I just think about all the tabs I have up right now and, uh, you know, the different types of information across those, the financial transactions where I need to verify my identity and things like that. 16:34 So hav- having the browser be integrated into your security posture versus a little bit of a black hole, uh, I think is, is certainly a necessary step, uh, and it's a place where we're seeing a lot of traction from enterprise customers right now. 16:45 You know, I, I'm sure you're probably asked this a lot, but, um, what do you see on the horizon for cybersecurity priorities in 2025? 16:54 Yeah, I think globally this is a real, I mean, really nice reset moment to recognize a, a lot of the progress and investment that has been made over the last few years, but redouble down in terms of focusing on cybersecurity outcomes, outcomes being that operative term. 17:09 Entities of all sizes, whether it's the Canadian government, whether it's the US government, whether it's a large bank, they rightfully spend and invest money in cybersecurity tools because they need to, right? 17:19 The resilience of their organization demands digital resilience. 17:23 I think y- y- we have seen this in public sector cu- all customers, but perhaps on the public sector side, you know, we shouldn't judge cybersecurity success by the inputs. 17:31 The success is not that you, just that you bought the tool. 17:34 The success is that you can show with real-time metrics that that tool is making you more secure from increasingly sophisticated adversaries like China or Russia. So outcome, outcome, outcome. 17:45 I mentioned some of those statistics at the beginning of our conversation, Jay, uh, about how sort of that mean time to respond, how quickly you can actually identify and then resolve a security incident going from two to three days to 14 minutes to two hours. 17:58 You know, those are the types of questions I think we should be demanding of all entities that deploy cybersecurity tools at scale. Like, it is great that you are buying and deploying this. 18:07 Show us how you, it is making you safer. 18:09 That rigor, that forcing function, the feet to the fire of metrics that prove cybersecurity outcomes, I think is gonna be better for everyone 'cause it keeps us all accountable, that we're not just focused on inputs, but we're actually focused on outputs. 18:21 We're a company in business to help secure things, uh, in a world of increasing vulnerability, which is why I'm excited, which is why I love my job. It's an inherently noble mission that gets me excited every day. 18:30 Like, we wanna be an outcome-focused company, right? 18:33 We make great products, world-class products, and, and there's high demand for those products, but we, we want a, a global conversation around driving outcomes, and we wanna be able to prove that. 18:42 So I think that is a theme that you're gonna see more attention to in the year ahead in capitals across the globe. You've been listening to Vanguard Radio, and we've covered a lot today. 18:52 We've talked about flipping the legacy paradigm when it comes to cybersecurity. 18:56 We've had input from one of the largest cybersecurity companies in the world, Palo Alto Network, and our special guest, Daniel Cruz, Vice President of Public Policy and Government Affairs with Palo Alto Networks, has delved into AI, secure by design. 19:14 We've talked about supply chain integrity and so much more, and it's been interesting, and the passion and, and commitment from Daniel and his company certainly comes through. 19:23 Daniel, I wanna thank you for joining us on Vanguard Radio today. I appreciate it. Thank you for the opportunity to have a, a really enjoyable conversation. These are important issues. I appreciate it. 19:38 The views, information, or opinions expressed during the Vanguard Weekly Podcast Series, CGE Weekly Podcast Series, Vanguard Radio Podcast, and CGE Radio Podcast are solely those of the individuals involved and do not represent those of Vanguard or any affiliated media products, their employees or owners. 19:57 Vanguard Magazine, Canadian Government Executive Magazine, and its affiliates are not responsible for and do not verify for accuracy any of the information contained in this podcast series. 20:07 This podcast may not be copied or distributed without express written approval of Promotive Communications Inc., PCI. You may not edit, modify, or redistribute this podcast without consent from PCI. 20:19 PCI assumes no liability for activities in connection with this podcast or for use of this podcast with any other website, computer, or playing device.