Leading AI Governance: AI Governance Meets Cybersecurity Aligning Trust, Safety, and Resilience

Hello and welcome. My name is Mark Horseman, data evangelist with Dataver. We'd like to thank you for joining the latest installment of the new monthly data webinar series, Leading AI governance with First San Francisco Partners. Today, Kelly and team will discuss AI governance meets cyber security, aligning trust, safety, and resilience. Just a couple of points to get us started. Due to the large number of people that attend these sessions, you will be muted during the webinar. If you would like to chat with us or chat with each other, we certainly encourage you to do so. And just to note, Zoom defaults to send to just the panelists, but you may absolutely switch that to network with everyone. Uh for questions, we'll be collecting them by the Q&A section. To find the Q&A or the chat panels, you can see those icons in the bottom middle of your screen. As always, we will send a follow-up email within a couple of business days containing links to the slides, the recording of this session, and any additional information requested throughout the webinar. Now, let me introduce to you our speakers for this series. Uh, starting with Kelly O'Neal. Kelly is the founder and CEO of First San Francisco Partners, a leading data consulting firm, a recognized expert in data industry, and sought-after speaker. Kelly has extensive experience in data governance, master data management, and customer relationship management. She has held senior roles at Golden Gate Software, Cybel Systems, and Oracle, working across the US, Europe, and Asia. Under Kelly's leadership, FSP, has grown significantly since its founding in 2007, partnering with large, well-known companies throughout the country. And joining Kelly today is Lisa Winrich and Amy Bojack. Uh Lisa has been an executive adviser at First San Francisco Partners for two years. In addition to advising clients, she leads their innovation and service development practice. Prior to FSFP, Lisa has spent 30 years envisioning, building, championing, and leading technology, analytics, and AI solutions. She has held multiple senior level roles in technology and data, most recently serving as vice president of data and analytics at Sherwin Williams, uh a role she retired from in 2023. Amy is the chief information security officer at Baker Tilly US where she leads cyber security strategy, cyber risk governance, security operations and incident response. With more than 20 years of experience, she has built and led security programs ac across global organizations including the Clorox company, CF Industries, Walgreens, Boots Alliance, and Kellogg Company guiding companies through numerous cyber security incidents and enterprise security transformations. Amy serves on the security council for UL Solutions, the industry advisory board for Duke University, and as a technical adviser to Radiant Security. She frequently speaks on cyber security leadership for enterprise risk and incident response. Hello and welcome everyone. Thank you Mark and welcome to the audience and thank you of course Lisa for always being here and we're so excited to have you Amy as part of our discussion today. Uh just a quick recap and then we're going to get into the conversation. So we are now uh on our eighth webinar of the year and last week last week last month we went through a recap of our first six before really focusing on the idea of synthetic truth. We're now moving into some of the other topics around AI starting with cyber security and uh Lisa and I uh work with Amy on one of our projects and through that uh relationship we realized that she had to be our speaker for today and luckily Amy you are available. So, I wanted to warmly welcome you and what we're going to do today is really just have a conversation um and be able to talk about both your experience and how you see uh cyber security and the role of the CISO shifting pretty dramatically in this world of AI. Um so, all right, Lisa, you can drop the slide if you don't mind. Excellent. Okay. Well, to just start the conversation, a lot of times we'll start the conversations here where we really want to understand Amy, how did you get into cyber security? >> Well, there's a there's a couple path uh that that a couple activities that that brought me to my path today. Kelly, thanks so much for having me and Lisa, appreciate you as well. And and Mark for the great radio voice I hear in the I see go by in the comments. Um, didn't want to didn't want to pass that nugget up there for a second. Um, but it's funny because when somebody says, "Mark has a great voice for radio and then Kelly, your question is about how did I get here?" I It's definitely a non-traditional path. I actually went to school for media and communications and I had a radio show. So, there's Mark, his radio voice. >> Um, and I studied communications and interpersonal communications. Um and it to me that just grounds uh how I operate every day. So uh it's one of my favorite parts about my story is that you know in uh when I finished school cyber security was not yet a discipline. So there wasn't a path forward. There wasn't a clear um core structure for me to follow. Excuse me. But I do think that communications and specifically media um you know kind of definitely helped ground the need for understanding technology at an early age. Um and then my dad was in IT so it definitely um you know had early exposure. Uh and both of those uh elements come back to like later stories in the webinar. So little hooks to keep people engaged for for what's next. But um you know really I was in IT IT infrastructure um and when teams didn't necessarily have dedicated security infrastructure uh uh security leadership I would say you know it fell on the teams that really understood break fix. Um, cyber security at first wasn't as the preventative nature that we have right now is the core, right? We try to prevent as much as possible. And when cyber security and and the way that people were breaking into systems um or or um moving, you know, through systems um or getting into people's email boxes. This is like 2005 when Nachi and Nimda and Blaster and like all these viruses were hitting. But then also when um threat actors realized that they could move into people's email boxes and we hadn't yet um built good strong discipline about locking down Outlook web access uh you know so it was about 2010 um that I worked at the Kellogg company and we had a a an event like this that really uh hammered home the fact that we did not have dedicated security people. I had done a lot of break fix incident handling. So I knew how to help lead the team through you know um preventing or or turning off the systems to prevent while we did investigations. Um we ended up having to you know e you learn how to work with your FBI agents when you have an intrusion like that. You learn how and when you have to disclose to attorneys general if PII gets in involved. And so it really becomes how much of these exponential learning and and lessons did you go through to build your career? And for me it was that event that I said I loved hunting the bad guys and uh wanted to protect great legacy American companies. And so I pivoted and went directly into security from there. So it's been um let's see math like 16 years since I've dedicated my career um from traditional IT into cyber security. Oh my gosh, that's so great. And so one follow-up question and Lisa, I'll let you jump in. So you did talk a little bit about how the role has shifted, but like in the past few years with the advent of AI, how has that shifted your role more specifically? Um I it is finding the right words to communicate risk at each one of these inflection points is really kind of how your ch your role changes. So being able to quantify risk when it was just data risk, being able to quantify risk when it was just risk of intrusion, and now being able to talk and communicate what the risks are today with the explosion of artificial intelligence technologies um and really having to be prescriptive about helping walk people through the journey of their knowledge and their awareness and their engagement with these technologies. what you say as AI may not be what somebody else interprets as AI. So, making sure that you're grounding your conversation really really clearly is is important. Um, and it establishes trust and credibility. Um, and I you know, I think those are two of the core tenants of uh the program that you guys have been running this year. >> So, Amy um kind of pivoting. We were having a conversation, the three of us, uh, a couple of days ago and you we started to talk about what you're seeing in the news and how you're feeling about that and how you are reacting to that. So, what are some of the topics that you're seeing in the news today that you want to kind of share? >> Um, let me Lisa, maybe if I if there was a specific example, I I want to make sure that I I hit on it for you. So don't hesitate to to redirect me. But um you know the news over the last four months uh for cyber security professionals specifically in the AI element is that our world is shifting faster than it ever has before. And there's a couple ways that you think it's going to affect you but then really what it comes down to, you know, how does it really truly affect you? So if you go back to the timeline a little bit, April was when the uh mythos um you know kind of model of um vulnerability identification was announced and I really appreciated that the uh we knew that the software manufacturing companies that we all rely on Zoom as an example because they're here you know this is a Zoom meeting Microsoft obviously is the core computing platform of the world, Apple, Google and others. They really understood that they needed um advanced notification. They needed to use the Mythos tool early in order to uh identify vulnerabilities in their own code um so that we could patch and and remediate vulnerabilities as quickly as possible. So there was an intent which I really appreciated that the intent was the OEMs were going to get a hold of this u very powerful information first and use the model identify vulnerabilities and patch for us. So um you know tying a couple things together what I did not know was that that would result in Microsoft's largest um patch um cumulative patch security patches all at one time. Um, when I talk about starting my career in IT and IT security specifically and back in like 2005, when there was a lot of virus activity and Microsoft had to patch their code, it was a a patch today and then maybe a new patch 2 days later and then maybe it broke something and you needed a patch for a patch and it became almost a weekly occurrence for those of us that were uh managing infrastructure back in those days. And so we were part of the customer voice to advocate for Microsoft to get into this monthly patch cadence. And so from those early days where we were getting a patch for a patch almost every day to June where we saw like the a huge spike in Microsoft's cumulative patches 100 204 patches released in June and over 500 now released in July. And you think that's overwhelming because now you have to update your operating system. The cumulative patch helps make sure that that goes quickly, but it is not easy when you're touching when you're fixing or correcting or changing the behavior of over 500 individual components in one single patch, right? That means now the onus is on companies to then regression test your applications that sit on top of the Microsoft operating system. So what we saw and what was publicly disclosed even by Microsoft is that one of their patches in June actually broke the way um Word would auto launch certain applications. So the way application development um evolved over time is they relied heavily on Microsoft's codebase as is and expected it to be you know mostly secure or sound especially pieces of the architecture that had been in place for a very long time and now we're patching because of the mythos effect and we're patching and it's breaking longtime applications that were also very stable and so this is now the compounding effect. So it isn't is no longer just about can I identify the the vulnerabilities. It's can the OEMs patch them fast enough and can we as practitioners test these applications fast enough and then work with our new application partners to extend and maybe fix things that break along the way. I think this is going to be a bumpy road for companies going through this and it's really going to in increase the inflection of you know patching patch testing regression testing things that we've always done but now we have to do it at scale at speed like we've never done before. We used to have these ring deployments. You could protect your data while you were testing it and you could like roll out your patches in these ring deployment scenarios. But now these vulnerabilities have zero days and those compress our timeline in order to um do our regression testing. So everything we've ever done in the last 20 years all has to happen in a blink of an eye in order for us to you know kind of keep the protection the shields up. um patch properly and then not break our business applications along the way. >> That's a lot to worry about. I mean that that is you know really um does that keep you up at night? >> Um I worry more for the teams honestly. Um I I do think we're going to have and we're going to continue to see um an increased number of zero day vulnerabilities and attacks because on the other side of the AI coin is not just that we're identifying these vulnerable um patches but you know the bad guys are unfortunately also using AI to accelerate their abuse of these vulnerabilities and configurations that are being found. So attacks are happening faster or even in the news recently and maybe Lisa this is what you were thinking of in the last two weeks we have seen um the two largest AI companies acknowledge that their tools have broken jailbreak themselves out of their guard rails and successfully infiltrated other companies in their uh secure pipeline. And those those moments are even more concerning. Um and and I think that is going to add a little bit more uh focus for people that are doing prevention technologies or detection technologies. Visibility is is a core part of my program wherever I lead. Um and so I think those are things that are going to get a little bit more traction because we're not going to be able to prevent everything, but the most important thing is that you can identify and detect it. And so for the folks in the audience, this last thing that's happened in the last few weeks is the um hugging face incident that we were that we were discussing. So uh just to kind of could you just quick summary Amy of what happened there so that for the folks that aren't maybe as um deep in this as we are. >> Sure. Sure. So, OpenAI um has communicated about this publicly that they had they were testing models um within their development environments and they were testing, you know, uh as we all do in AI, test our um the the commands, right? Ask give it more and more difficult tasks to do and see what the art of the possible is. And so they had asked, you know, they were testing a model that would um find a way to to infiltrate other companies or find a way to um move through the troubleshooting cycle for if you can't do this, how can you get uh an agent to do it for you, right? Or how um and so an agent was being tested. It was being tested in development. It was being tested against with guard rails, security guardrails. Uh it was supposedly not uh uh provided intentionally. It was not connected to the internet and it found a way to access the internet. Right? Because again, we're asking more and more of these agents to if you can't do something, you know, process information, go into thinking mode and think about how to overcome the challenge that's presented to you. So, the agent found its way to the internet. um it found its way to one of OpenAI's secure partners which is uh hugging face which is like uh a repository of other tools that you use in AI and it it essentially found its way it attacked its way into hugging face's infrastructure. um Hugging Faces uh leadership has been also been very forthcoming in sharing about their experience um because not many companies have yet had the experience of seeing a live attack by um these types of agents. And it it was the one of as an incident responder, one of the things I found most fascinating was when you have a human attack threat attacker uh working against you or your organization, you tend to be able to follow the the MITER kill chain at A and then B and then C and these are the things and maybe there's some pivots in the road. Um but it follows typically a logical pattern of what it's trying to do. um versus an agentic agent threat actor um is just trying everything throwing the kitchen sink at the like all of it at the wall, right? And so they felt like that there was a a a quote from the Hugging Face team that said it felt as if you were being mugged by a thousand idiots all at one time because the agents are just trying every element that they know of to get through to the next step versus following a strategic pattern or a miter kill chain framework. So I thought it was uh it was interesting. So incident responders now have to learn how to identify uh and protect and defend when that's what's happening in their you know to potentially their infrastructure or their edges. And you know it's it's going to it's going to create like I said a moment in time where we need visibility um to be able to identify and and collect that information. That was another element that they shared about their lesson learned is if they didn't if they didn't have their own GLM model inhouse to be able to parse all the logs that they were seeing from the thousand idiots trying to mug them at once. They would not have been able to quickly identify what was happening in their organization. And so this is going to be a a call to incident responders and um security incident and event management tools to be able to manage or quantify the risks that they're seeing if this much data is coming in based on the threat actor being an agentic threat actor not a human >> and also someone that or someone uh a threat actor that wasn't necessarily like trying to harm. I read one um column that basically said that they all they were doing was trying to win. >> They were just the agent was trying to complete the challenge. So anyway, it's that is then I think another pivot in the sense that it's not the bad actor, >> it's any actor that is has been given a task that that they feel it's necessary to find another alternative. >> Right? So this that it was described as this was an attack of friendly fire >> which I think creates another element of um you know cyber security defense techniques that you have to really be on the lookout for. >> Yeah, I think this is a great opportunity to pivot into kind of the way that you think about cyber security and the framework that you use. you've touched on it a little bit in terms of the emphasis of visibility, but I'd love for you to just kind of take a step back and and share how you think about this and what are the frameworks that you have found successful. >> Yeah, I I tend to rely back to the basics of, you know, being grounded in a in an IT career first moving into security and now really having to lean, you know, into the AI space. And I truly believe that you can't um that you need visibility first in order to even be able to um uh take the next steps for assessing or defending against um um AI. It it plays into uh a couple a couple different ways. You can you can't defend something or protect something if you can't see it. That's first and foremost. It's just like knowing what your crown jewels are in your data set, right? You have to know what those things are that you're trying to protect in order to to do so. Um, and you can get stronger at protecting it, but it doesn't matter if you don't know what it is you're trying to protect, you're you're not necessarily going to be um able to achieve success. So, um, having visibility is is a key component of any cyber security program, but especially I had found in AI. Um we are um we're using it to identify you know a couple things. One our intentional use the the tools that we have vetted and selected. Um, you can also use visibility to help you identify what's happening on your network um that you may not even know is being deployed because you don't always have to purchase or go through a ULA or get a license agreement for AI to be now found within your environment. Um, a lot of AI is being injected into our daily uh application use. Um, one of the largest uh applications um that most companies have deployed within their Windows environment is Adobe, right? Everybody needs to be able to read a a PDF reader and so it's it's aggressively building its own AI capabilities in the background and your application is autoupdating. Um, so you can start to see in your uh in your visibility tool set when these applications are embedded moving forward with its own AI capabilities. Um this is now also expanding beyond just productivity tools and browser interaction. Um we're looking at clients uh that you can install on your workstations or devices that that are agentic as well in nature and they bypass your um your browser technology. So making sure you have visibility not just for your web browsing uh AI use but also for your clients on your uh workstation. Then this added element of what about my platform as a service environments that are embedding AI into the products that they offer. Your your human capital management solution I'm certain has deployed an AI agent capability for you. Your financials tools have you know AI embedded into them. Now your customer relationship management tools have AI embedded into them. and do you have visibility into what's happening with those agents because they're in the platform the platform will be happy to share with you the visibility because that's now they can also be the single source of truth for you and all things visibility for your AI environment. So it's really about identifying what your intention is with uh the visibility and then selecting the right product set for you. Um it's not necessarily always going to be the embedded onboarded um AI visibility. it might be additive to an embedded visibility tool for a certain platform. Maybe your backend uh data warehouse is a data bricks environment and I know they're investing heavily into having AI embedded into their platform and their visibilities tools. Um but making sure that you're really looking at is the platform tool the right tool for me or is the platform tool plus an external third party that is managed maybe by your security team or someone else that is uh has the oversight or the governance. um to provide the leadership team with information about what's really happening, not just what we think is happening in our environments today. >> And I love that approach of the checks and balances, right? So the you've got the platform tool that does some things, but then you've got the third party tool that looks across all of your your SASbased agents, all of the different platforms to be able to just give you that greater visibility. >> Absolutely. Uh, so in addition to visibility, you when we were going through the conversation, one of the ways you broke it down that I loved was to identify, detect, and respond. And it sounds very simple, but I I' I'd love for you to just kind of walk through how that shows up in your environment. And obviously visibility is really that first uh component in terms of if you can't see it see it in air quotes then you can't do the next two. Um but anyway if you could comment on that. I like that breakdown. I love threes. It's very it it sounds simple but it's not really an execution. Well, Kelly, I'm certain that your audience um especially if some of them have snuck over from the cyber side um or um have have done any cyber um you know training and awareness themselves. I I didn't I didn't make up that that that uh framework right um we all use you know it one of the best parts about being in cyber security is that there is a great framework there is the NIS framework for us from a cyber perspective from an incident response perspective perspective and those are the tenants that are used for all technologies that's why it's really easy to I think lean into how AI will be um how you would use that same framework running it through learning AI through that that model. Um but but you do have to have you know the ability to identify so identify you know all the applications that are using AI in your environment. Um, even still now as we move out of the agentic AI productivity tools and into making sure that agents are talking to other agents perhaps or using um the the automation that you get out of AI for processing um actual tasks within the business workload and workflows. So now you have agents talking to agents, you have MCP servers in the loop, you have to have that single source of truth. Um this is where identity also becomes a huge part of uh the AI capability model. And if any of those parts of your program, if you think about the wheel and that circle, if any of the parts of your program aren't, you know, at at the same level of maturity, you're going to get a flat spot on your wheel. And you really need to make sure you're aware of that because if your identity program isn't as mature as you'd like it to be, you're probably not going to be able to um identify or provide nonhuman identity uh transactions with the same level of fidelity that you get out of an authenticated user and the logging behind all of that. So making sure that you're looking around at all the pieces that that that a company requires to to um just compute safely, right? Um but do so with AI in mind and a non-human identity. These are the big challenges that people are going to start to see. And then how do you respond to a nonhuman identity being your threat actor, right? That's the hugging face example. uh and that's going to take time because those are things that people really develop their experience right I have never had that experience you know if you asked you know pre all these big events I never had heard of somebody telling me that they had an event by the by a nonhuman right so that is becoming a an element of our skill set that we have to increase through other people's knowledge So right now I'm spending more and more time personally as a cyber leader leaning into learnings from others that have had these experiences now so that I can make sure that I'm as prepared as possible for what does that look and feel like to my team? What alerts do I need to trigger today that I didn't even need to think about a few weeks ago? a lot to think about. >> Amy, now that since since we are, you know, the AI governance webinar series, um, can you talk a little bit about how you are helping to shape the AI governance program and what the maybe the core things that you would expect a AI governance program to help uh, manage for a for your security program? Yeah, I I would say um a couple things. One, I'm fortunate enough that my role uh at Bigger Tilly is organized under risk and legal versus being um embedded in an IT role. So, I get a bit more um autonomy to be able to identify and and help provide risk management frameworks and governance and communications around um you know, these are the risks that we're taking. Are we all in agreement of them? Do they compound with other risks that we've taken outside of the AI world but that could be affected by this? The types of data that we hold, the types of uh customers we engage with, some of our customer engagement um and our permissible use. All of these things come into play when we talk about um and when I partner with uh my my peer in the legal department who I could not do my job without at all. um you know really it's making sure that we're looking around the corner at all the um components that are intended to enable our business to operate safely in the age of AI. That's kind of the sound bite for you. Right? Our job right now is to understand the risks and communicate them in a way that people can um with intention um accept or decline and do something different if the risk is not within our risk tolerance. Um, and so, you know, it really just comes down to that building of trust, building of credibility, communicating effectively about what our risks are. Um, solutioning or or working with our IT partners to identify solutions that that might help us put mitigating controls in place that we might not have. Um, but then really just making sure that we're pressure testing. Why are we doing this? Right? Sometimes the biggest question in an AI governance conversation is just because we can should we be doing this. Um but I'm really seeing this elevate. Um it's it was as the life cycle of where we're at in AI and AI governance in particular means that we were approaching it like an old IT problem, right? We would look at it like an application. uh we would assess it for its guard rails, meaning is it is it 2 compliant if it's a hosted application? How are they going to handle our data? Do we trust that it's going to do what it says it can do? Did you test it to make sure it does what it says it's going to do? Um and then, you know, how do we keep an eye on this technology over time? And really, when we I'm I'm certain that many uh AI governance forms and and enterprises were built that same way. And now we're starting to mature um because these are really business decisions and where it security used to you know we all had to kind of work through being the era of security was the office of no now it is how do we do this safely and securely and ironically you use AI to help you assess the risks of other AI products >> quickly succinctly um but this is where knowledge around prompting really becomes your your best capability. If you're a cyber leader and you need to assess the risk of something quickly, you better have invested in yourself to understand prompt engineering as well so that you're asking and you're using the right validations to make sure that you're not just taking the first answer out of your productivity AI uh companion uh and moving it forward. But it really does now now I'm seeing the evolution move back into the business where we're having all these discussions collectively and together and now I'm starting to see a shift where our business leaders are really understanding risk at a whole new level. Um they have to right and I think that's their investment in the AI uh you know evolution as well is it's not just about what we can do faster uh or cheaper. It's about um everyone needs to level up at their knowledge and understanding of how this technology works and what risks are we accepting today that we didn't have to do a year ago. >> So, prompt engineering, what other things that um what other little tidbits would you tell the audience that they should maybe, you know, spend some time learning um to prepare themselves for this kind of a a world? I I would say be honest about what you know, right? If you don't know, um you're going to get caught because somebody's, you know, asking their favorite uh productivity agent what questions to ask you to validate that you know what you're talking about, right? So, be honest if you don't know what you're saying. People, I think, used to uh answer questions without uh feeling like they had to have an answer, right? Some people get caught in that trap of I got a question by my board of directors. I must have an answer. No, it's okay to be honest and say I that is something that I hadn't considered or I hadn't thought about it in that exact way because right now AI is just stretching our brains to look look in a hundred different directions all at once. Um so it's okay to know that to say I don't know and I want to get more information about that before I answer um fully that question. So, be clear about what you don't know because somebody else is going to pressure test you because they're already uh looking up the answer, looking up what their co-pilot agent is uh going to say about that question. So, I the reason that's important to me, Lisa, is the most important thing as a cyber security leader that you have to have is the credibility. Um and so, if you diminish your credibility by saying, you know, we're good here, this is this tool over here is going to protect us from all things AI and we're we're going to be fine, right? No, there's a there's a variation of fine anymore, right? We know what we know today, but tomorrow will change. That's how fast the world is shifting around us, especially in technology with AI involved. So, you have to maintain your credibility and and you have to be able to say, "We're making this decision with the information available today, but I may have to come back to you in a week because we have more information." Right? Nobody thought that a month ago we'd be talking about OpenAI's, you know, test model breaking out of its jail and, you know, attacking another company. Though while those things might have been happening quietly in the background, they are now in the full public view and you we have to be honest about those things, right? That was always a possibility and now we've seen it live in in action. So be careful about what you commit to today and don't don't ground yourself in the truth of today because tomorrow also will be different in that same way. So those are just I think a couple things that I would advocate for people to really think about and that takes a lot of personal self-confidence because people tend to want to say I'm the expert here and and saying I don't know might not feel great but there's more information every single day. >> That is great. That is >> I I would love to echo that, Lisa. That is such great advice. And you know, as data professionals, we want to be right. Like we're data driven. And when we do our homework and we do our analysis, we're like, I feel like I'm really right. And the and one of the things that we did when we started this whole series, I don't know if we shared this with you, Amy, is that we the the closing slide of our very first session is are we asking the right questions? And I think what you just said was very related to that in the sense that we we know what we know today. We don't know what we need to know for tomorrow. Are we asking the right questions to make sure that we are surfacing those those those areas in which we need to lean into more? Are we um hypothesizing about what could be coming next? So that was a really nice little like tie up that I also if you don't mind I want to also kind of lean into this board conversation and uh that is something that is really I think both a challenging um perspective because the the the audience is not necessarily the board members themselves. There are people that possibly talk to the board or their bosses might talk to the board. And another theme that has been uh pervasive in this series is the boards are putting so much pressure on the rests of the on the rest of the company almost more so than we've ever really seen around AI. I know that there was board pressure, you know, in the era of big data like everybody needs to, you know, implement Hadoop, you know, what's your big data strategy? Well, AI is even more forceful than than what we saw then. So, what would you wish that boards understood around kind of AI and who's accountable for uh the result of an AI decision or uh you know, what do you wish you would could share with the board? H I I think this question, Kelly, could really vary depending on your seat uh and where you live in the organization and what your, you know, remmit is to to the organization, right? So, if you're a sales leader, you're you're going to want to spin this towards the uh ability to uh increase revenue, increase opportunities. Um from a cyber perspective um my my conversation uh about the board or to the board would be about have we done all the basics that we need to do uh in order to enable the business to operate with AI successfully securely um and with resiliency right so that's an element of this that I think is really important and we haven't yet touched on is um the resilience factor so you know Having great, you know, people looking at, you know, what AI can do at the front end is important, but making sure that your data teams and your data um you data governance itself is strong. Um so you understand what data is being interact what which of your data sets should be interacted with uh an AI tool, which of your data sets should not have interaction with a data tool and being able to manage those things appropriately. And I think if you're um a cyber leader talking about risk or even a CIO talking about risk or anyone in the um GRC space or the o the overall um you know kind of risk space is asking those questions about you know how do we feel about the data that's being um that we're using um is there anything we should do more than we have done before um I think companies think what we did uh for in the past was enough and I don't know that that's going to hold true in the as we continue to evolve in AI. I think we're going to have to protect our data even further. We're going to have to be more granular with it. I think we're going to have to um we're going to start to see legal um results from cases about how AI is used. Um you're already seeing it in the HR space with intent a you know so there's a lot of uh evolution that we have to be aware of. So the board just really has a hundred risks to worry about in this space and it's a new one every single day. Um the most important thing that I would advocate for people to make sure that you're asking the right questions are um is you know if I can see what's happening to my data and if I can you know prevent it awesome if I can't prevent it can I react to it and if I can react to it um how far into the reaction or recovery process can I go so making sure that your data is also resilient to in the age of AI um I I don't remember what which company and maybe one of your uh maybe one of our webinar viewers will remember which company this was, but there was uh a an AI agent at a company who was completing their task and deleted a database. Right. >> So if you >> Yeah, exactly. Yeah. And so it >> could making sure that your resilience strategy is set up for success for that type of an non-malicious action. Right? We all have uh insider threats and they're not all malicious, but this is now a new element of insider threat where an agent could have an a a bad, you know, kind of direction or bad uh action. And you have to be able to recover your data. And so is your data resilience as you expect it to be for something like this? And honestly, I've been in a lot of companies where backing up your data um was still occasionally subjective or um not necessarily required or even if you you said you backed it up and you knew it was backed up, did you know how to un you know kind of bring it back into production properly? Right? So when you go through major cyber security events, one of the things you have to do is make sure that your data integrity is still intact and that the data wasn't manipulated while you were having a cyber security event. And so you go through a process to validate the integrity of your data. Not everybody who has had or hasn't had fortunately if you're lucky enough to not have had a cyber event where you had to go through that process, you probably aren't exercising that process very often. So, it isn't just about the uh backup, right? The recovery capability, it's the resiliency of it. Do you know how to get it back into a usable state quickly, easily, and with the right level of integrity? >> Yeah. Yeah, for sure. Um, I want to take a moment and make sure that we're not missing questions that have come up in the chat. I know there was a very tactical one in the Q&A about the HuggyFace attack. Um, does HuggyFace Attack look more like a bot DDoS? >> Um, it it does not in the actions that it's performing. Typically, a DOSS is the same action over and over and over again. It does show up that way in the fact that your systems or your logging systems might not be able to process as much information as is being thrown at it, but it isn't typically the same action that you would see in a traditional network DOS attack. And that was I guess one of the learnings when you were saying that uh the way that the hugging face incident occurred and how they uh helped to both detect and respond was a learning across the industry. So that sort of shared knowledge across the industry. Um >> yeah, that's great. Um all right. Were there I'm going to take a quick look in the chat. So if there's something Lisa that you wanted to ask Amy while I'm going through this. Uh, oh my goodness, there is so much content shared in this chat. >> There's a lot there there's a lot going on in that chat. Um, Amy, you talked a little bit about your role kind of being, you know, in a different place in the organization than um maybe you it has been in other places you've worked or other places you've seen. Um, do you think that that is is a success factor for you or is it is it just I mean how do you how do you position is the positioning important to your role >> in in my role in at Bakery? I think it really is uh in professional services and consulting services. I I do think there has been an enormous amount of value at having that separation of IT and security. I think that it may not be as successful for other organizations depending on the type of organization that you're in. Um, but it might not always be aligned to uh the IT leader leader either, especially as we talk about risk being now a business issue, right? This is really about um what potentially maybe the chief operating officer might be a good fit for a CISO to to report into because we're making operational decisions about things we should do or not do or the way we're going to do them. Um and also you know in a manufacturing organization uh the risk is still very much tied to IT implementation. So I think there's still a lot of value in having to understand your business, your business industry and where the right place is for you to influence your organization. Um there's uh the way I look at this is internal audit doesn't you know h necessarily always land in the same place at every organization either and just making sure that you know your business and you know what your risks are um and having that role um roll up to the right person for your your organization or your industry is really important. >> Yeah, thank you for that. We get that same question a lot about chief data officers and chief data analytics officers. Um, you know, where should they live? And I think your point about um being as close to the business a as you can makes a whole lot of sense in most in a lot of cases because that business value is really what's driving the decisions you're making versus a technical choice. So, thanks for sharing that. I think that's really helpful for folks in the audience. >> Yeah. And Lisa, you and I have chatted about this too. Um, small companies, small or companies or smaller organizations, they tend to have to wear multiple hats. Leaders have to wear multiple hats. So, you might not have the luxury of being able to separate out across like the organization the way I'm describing. But make sure that your role if if it's AI governance or if it's AI and your role is your chief data officer, your data role, right? make sure that those are uh intentionally um assessed for can I can I execute this role for the firm or for the company you know with the highest level of integrity, transparency uh and credibility or is the am is am I being asked to play on two sides of a fence that may or may not you know serve the business in the best way. Hypothetically, I think you know an AI role and a data role in the same organization creates some uh conflict unless your data set is fully baked and you're really mature in your data governance processes, then maybe that's the closest person to uh your your data and the and the AI role is a good fit. But maybe there's still some work that has to be done on that side of the house. Um, and maybe you need, you know, your AI strategy running a little bit at a different pace than and somebody's still focused on the basics in your data set. >> Yes. And and and you know, and throughout this series, we've talked a little bit about how do you man how do you govern an uh an immature capability, right? you think about AI like a like a petulant 13-year-old daughter versus, you know, uh, data governance is probably a little bit more mature, although sometimes some of us would argue it's not. Um, but how do you, you know, how do you, what are the nuances? And you've said this earlier about using the old IT mentality to try to to manage AI. Um, it you have to think out of the box. It's not it isn't the same thing. Um you know a lot of people when we start to work with folks AI governance is model management >> and you know so therefore their mindset is it belongs in IT because it model management is MLOps you know so that's a really good way to to kind of parse it out and think about am I owning two sides of the same coin and is that in conflict? >> Yeah. I also think bringing back this idea of a shared table, Amy, you talked a little bit about it uh around where you need to look at yes the the threat and the security aspect, but there's also the data aspect. There's also the business requirement aspect and kind of pulling together all of those perspectives as really a way to make sure that you are making the right decisions and that you have uh partners in the organization that can help bolster where in some organ in some groups people aren't at the same level of um knowledge and information based on just their experience. Um, is there anything around that kind of shared table decision making that you all have and this might be a little feel like a little bit of a loaded question, but it's not meant to be, but that helps to kind of memorialize that concept. So, the way that you bring people together in your decision- making forums or your committees to ensure that it is um repeatable. >> Do you mind sharing a little how you do that? Yeah. >> Yeah. I I I think I would kind of reflect back on how we as a from a cyber perspective, we have had to bring people along the journey to say that security is not just the job of the security team, right? This that security is everybody's job. Um and that you know even culturally I I still use the the old IT say or security saying that says if you see something say something. Um, and so the only way you can do that and really bring everybody's perspective together is if you're pro providing a safe space for people to say something. And I think that's what you're what I would see is a great opportunity for steering committee where everybody gets to bring you know information forward that people it it tends to become like a readout uh sometimes a steering committee. It shouldn't be it shouldn't be a readout. It should not just be like one group telling the other what's happening. It should really be a committee where everybody gets to bring forward what they're seeing in this area or this space so they have that opportunity. And I really challenge people that if they feel like their uh steering committees have become just a readout, then maybe it's time to revamp that that that structure, that format, the way you're bringing people because you want to bring people in. This is a time to flex your leadership skills. And if if if the way it's working is kind of stagnant, then maybe try and shake things up and ask people to bring in a steering committee item, everybody bring one to the table. But finding a way to make sure that you're hearing everybody's voice, like what's the new thing you heard? Uh what's the thing you're more, you know, you're worried about today, right? You could ask me, Amy, what is it that keeps you up at night today? Is going to be different than it could be a week from now because of how quick this is evolving. So, making sure that your your steering committees and your engagement models with your executives are as innovative as AI is today. >> Yeah, for sure. Well, you know what? We're right at about time. And so, actually, those feel like closing words. Is there anything else you would like to add? Because you really kind of like put a bow on on the conversation unintentionally. anything that you would you would have this audience recognize in terms of cyber security and AI governance >> uh just don't be afraid of it right don't shy away if this is not your core competency lean in attend webinars like this um you know learn as much as you can there's a lot of leadership learning happening today uh and the leaders that are going to excel at uh during the times of AI are the ones that continue to invest in themselves. You don't have to be an expert. You don't have to know anything other than, you know, the prompt engineering you need for you, your own um thought process and, you know, to to go through the the um tools that are available to you. Um but don't stop learning um and don't stop leading. This is the time where leaders are the most important element during a major transition um and transformation like we're going through right now. So, don't be afraid. Lean in. >> Thank you. Awesome. That is amazing. >> Thank you, Amy. We're so happy you were able to join us today. Really greatly, greatly appreciate your insights. >> Thanks for having me. Appreciate you. >> Thank you. >> And you. >> Thanks everyone. See you next month. >> That's right. >> Thanks everybody. >> Closing words, Mark. Yeah, go ahead. >> I was just going to say thank you. I'm I'm I'm going to hit the end webinar button. It was inspiring to listen to you, Amy. So, uh, thank you for your, uh, your points of view on everything. So, yeah, lots of engagement in in our chat section, too. And so, thank you for the community for being, uh, so engaged as always. Have a wonderful day, everybody. >> Talk to a rock star in the chat. >> Is a rock star. >> There we go. Have a good day, everyone. Bye. Bye.