Transcript of https://www.youtube.com/watch?v=5EESmMZ-PY4, published 2024-12-10. Auto-generated captions, lightly corrected; the video is the canonical version.

AI & Cybersecurity Compliance: Key Strategies to Protect Your Business

I'll go I'll go first Jim Goldman co-founder of Trava security um 25 or so years in cyber risk management um experience with uh FBI cyber crime investigation U vice president of GRC for Salesforce and really just a fan of risk man cyber risk management in general so when ISO 420001 and the nist aai risk management framework and the eui act came out I just dove in head first and have started to help our our customers with how do you do you know risk assessment around AI awesome hey everyone Mark ranaco here I have about 12 years of experience in cyber security kind of kick things off um in like a data center incident response role

that got me really interested in in red teaming penetration testing vulnerability scanning that's where next role was and then I eventually got into cyber security compliance was a PCI DSS auditor for quite some time audited against some other uh Frameworks and acronyms that I'm sure you all heard of and now I've been with Secureframe for about four years Building Product to help organizations manage their their GRC um and over the last couple of years AI has gotten big I I've been uh working with a lot of our customers now that have ai implemented in their solutions to help them try and figure out a way to the Implement risk management and and continuously monitor that

that risk management which we'll be talking about uh during this webinar today so with all of our webinars we like to make it very practical we want people to leave the webinars actually knowing uh to some degree at least what next steps they would take what concrete next steps they would take we don't want it to feel like a college lecture but rather like hey here's some practical information that you can put to good use right away so for context we're going to start by talking about AI risks in general and how to manage those we're going to talk about the standards that are out there to help us in regards to this and then

we're going to get into more of a process thing like how from a practical standpoint how would you get started and then why would compliance to one or more of these standards help you as a competitive advantage and this is for you mark go ahead the next slide sounds good Jim kind of kick us off and basically set the stage um you know as technology starts being implemented into society we we do see a a a rush to or race to try and Implement a lot of that Tech nobody wants to uh to miss the boat on on the next new greatest thing and we're kind of seeing that in in the realm of AI

over the last couple of years um in terms of this like AI Revolution that's happening uh organizations are implementing AI into the software into the platforms that that they've already previously built we even see organizations starting new companies based off of AI using generative AI to try and solve a new problem um and if even if the organization is not implementing AI directly into their software a lot of the users within that or are using AI to try and enhance or or make more efficient their their daily tasks with work um and and that's all kind of what's happened over the last year is all these organizations are rushing to to implement AI to make

sure that they're they're using it to their best of their ability and and they're they're keeping up with with the new technology uh and because of that um we're we're we're not really seeing a lot of regulatory uh like standards or or practices today in terms of the use of that AI it's kind of like the Wild West where uh organizations are spitting up AI for for any service that they could possibly think of starting new companies onboarding users taking a lot of data and processing that data and there aren't really any sort of um requirements around you know what you can and cannot do with AI but uh as we'll be talking about in

our next few slides that that is going to be changing and there are Frameworks and things like that for for you to be able to kind of kick off that process to determine if you're using AI in a in a ethical and effective way um so basically you know due to the rush to implement AI a lot of cyber security and and privacy considerations are being overlooked um and and proper due diligence in terms of how to implement Ai and process data within those systems are are being overlooked as well and the reason for that is kind of generative AI organizations make it so easy to implement it into software if I wanted to change

one of the functions in my application to start using Genai all I really have to do is connect an API and then start feeding that data to to the generative AI tool which will then just begin sending results back and then users at my org can can log into a website start asking a generative AI some questions about the business processes and it'll spin out results or or content that can be used to help make that that person's job more efficient so it's just it's really making it's very easy to implement and because of that um harder harder to regulate versus you know slowing down the implementation um and a big issue with that is

the results that are coming back from from these generative AI tools can't be 100% Cur it or trusted upon and the reason for that is because a lot of these results are are trained on previously used data or are based on the prompts that are being sent to these tools and if there really isn't a lot of due diligence or um you know research around how the AI should be used in the first place the the results that are coming back could be untrained inaccurate unfair and biased uh which you know could cause a lot of issues in the long run for these organizations and not only that a lot of the data with these

gen tools is not continuously updated so a lot of the data that it's being trained on and the results you're being given is outdated inherently and then how that AI kind of makes decisions for or the data that's being sent back uh it's it's hard to see exactly how those decisions are being made uh these these tools are not very transparent in how it's taking the prompted data turning it into a result or an answer um and that makes it really difficult to to kind to understand you know is the data coming back accurate how do I determine if that if that's so um so be because of that um and here's a graphic that

that kind of shows you um all of the different specs and and how it all relates to data and and Ai and I'll kind of let Jim talk about the graphic a little bit here yeah thank you Mar so it is that rush to AI that Mark referred to that really is the root cause of all the risks that are now apparent or inherent in the AI systems that are being developed and what you see here and Mark was referring to this is any AI system really starts with data data is the foundation of these and if the companies that uh maintain that data Lake don't have the good data management best practices and there

are standards out there regarding that that we won't go into today but if they don't have data management best practices with data cataloging you know eliminating data redundancy concern about data quality Etc then it's almost like don't pass go don't collect $200 full stop don't move up the stack you can't skip any of these layers right yes AI effective AI trustworthy and effective AI is at the top of the stack but understand it is a stack and none of these steps can be skipped you can't skip over the data privacy regulations that are required you can't skip over the cyber security standards and you have to use you know best practices in terms of AI

systems development so forth the the kind of the point that people Miss Even if they're well- intended and say well I want to develop an AI system that is um in line with ISO 420001 or in line with the nist AI risk management framework you can't just jump to that you have to start back down at the data with your data management best practices data privacy regulations cyber security standard and then you're ready to layer on best practices in terms of AI systems development that's really if if you get no other message from this webinar that's in my opinion at least that's the most critical message to bring home from this and so this is

kind of you know best case worst case scenario so if you were to uh follow that diagram that we just showed and you started with you know good data management practices then these are the characteristics of the system that you'll end up with you'll be compliant with regulations you'll be able to prevent data tampering you know your system will be trusted you know you'll be able to mitigate attacks etc etc and then uh without going through that stack without starting with data management and incorporating cyber security and D data privacy regulations you know you have damage to your brand and reputation you know with the EU AI act which we're going to talk about in

a little bit you'll end up with some pretty significant fines uh from the EU I I'm I imagine you know there's a California AI act that's in the works just like California LED the way with privacy regulations in the United States and other states followed I'm sure there will be other AI type acts you know with um punitive um you know penalties as a as a result of not failing to meet standards uh so you know it's pretty pretty straightforward it's in everybody's best interest to follow that diagram that we said start with data management include data privacy include best practices in terms of cyber security and then follow the safe Frameworks for you know

um comprehensive AI system development did you have anything to add Mark on that one yeah I think you I think you're covered to Jim basically like uh as we talked about in the graphic before and here I mean there there's a lot of risks in regards to the implementation of AI and we'll be talking about how to manage those risks but a lot of those risks could also be helped mitigated through a lot of fundamental controls through privacy and cyber security regulations that you may already be following and maybe you just don't have your AI system kind of involved in those best practices so um so yeah these this is a list of off the

top of our head just like a bunch of different risks that that you should consider as you implement and and use AI so in terms of risk assessment you know you can't fix what you're not aware of right and so uh the first step in risk management is a risk assessment you have to know well what are the risks I'm dealing with in my particular ular situation for the AI system that I'm trying to develop and so when you're trying to do an assessment a risk assessment you have to have some kind of gauge to measure those risks against you have to have a framework you have to have a set of best practices that

you can you know kind of go down that list of best practice and say are we doing this at all or maybe a little G we're not doing that at all etc etc you need to be able to work your way down and compare those kind of best practices to what you're doing so that's what we have over on the L there is you know you need to choose which standard or framework uh is most appropriate for you Mark's going to go into these in a little bit more detail later so I won't do that now the one thing I would point out is ISO 420001 um is obviously an international standard that you can

get certified to the nist AI risk management framework is just that it's a framework of best practices no certification but it is very similar to ISO and that you know it says in our opinion this is the best way to do this the eui ACT is different all right the eui ACT is law it says Thou shalt do these things and if you don't do these things here's what's going to happen to you so please understand that the eui ACT is legislation yes it it puts um out some best practices many of which truthfully came from ISO 420001 or the N uh risk management framework but understand that it is legislation and and so um

you know how do you go about this the risk assessment well you you have it starts with processes and and really it starts at the very top what are you trying to accomplish what's the use case and what's your business justification once you do that you really look at AI systems development as one long process you can think of it as like a manufacturing process a quality management process remember the first ISO standard that really became well known at least in my opinion was ISO 91 you know which is a quality management standard and so really this is you know picking up right with that left off it's how do you develop AI systems in

a qual quality manner well the first thing you need to do is what's the overall development process look like and then what are the roles who can touch that systems development process at any point in the process you need to know what those people are and then you need to be able to define the roles and responsibilities of each of those individuals in that process so just understand we're trying to build a process that a quality process not only to develop that AI system but to deploy that AI system and then to be able to monitor that AI system once it's out there and operational and then and then you you know really in the

monitoring part what you have to be able to understand is what does good look like you know you have to build note exceptions no you have to have um thresholds set for different types of activity so that you know when you're kind of out of bounds on those parameters you know and I I'll just say one other thing related to that last category about um unacceptable high or Limited in terms of the use of the AI those terms again come into play with the EU act because again because it's legislation it's it comes right out and says Thou shalt not Implement these types of functions in AI systems and then it says okay if you

want to do these functions it's okay but we consider those high risks and therefore you need to implement these extra controls for those types of AI functionality so just be aware you know risks fall into about three major categories U pretty self-explanatory you know there's the operational risks bias is probably the biggest one you read about bias and fairness because you have to understand this AI system is giving you answers it's making recommendations all right there's a lot of liability in having a computer make recommendations and so you have to be very careful on the fairness and bias scale uh you know the regulatory risk come into do you even have the right to be

using the data that you're basing your recommendations on and then the second thing is um transparency is a big deal I think we may have called it traceability that kind of thing again the eui ACT is very specific about logging requirements like you need to know what you know every recom every question that was asked to your system every recommendation that was made you know how'd you reach that recommendation etc etc so this is not a thing where you can just throw an AI if you're going to follow the standards and avoid risk this is not a thing where you can just throw an AI system out there and start giving recommendations to people and

think you're going to be okay um and obviously the reputational risk is self-explanatory and then in terms of impact you know and it um I think everybody knows that risk equals impact times probability right and and so you know that that's really how you want to gauge what are your critical risk what are your high risk what are your medium risk what are your you know low or marginal risk Etc so what could the impact be you know what are the potential impacts um that could happen well certainly it'd be Financial again uh if you fall under regulation and you go sideways against the regulation then understand there will be Financial impacts to your Enterprise

um you know probably the next big thing is reputational right and then the operational impact comes in in that do you do you have the processes in place to monitor your AI systems know when it's out of bounds do you have a way to handle the data privacy kind of inquiries and that type of thing and you know how do you manage it well it starts as we said with the risk assessment but then really once you have the assessment you know where your risks are especially with ISO 420001 for instance you have to a very comprehensive risk mitigation or risk management program in place to say yes we identified these risks here's the our

plan on how to mitigate those here's how we're monitoring that risk mitigation plan how we're tracking the progress on the things that we promise we do so it's like actually do the things you said you were going to do document that Etc that's good risk management you know the other big thing with AI systems is you have to use threat modeling sometimes s people call it killchain analysis that type of thing where you you need to put on your bad guy hat you may want to employ red teams or you know I think there's going to be a new flavor penetration testing for AI systems where it's not necessarily penetration testing just in the classic

sense of trying to break in and steal data but it's like can we unduly influence the recommendations that the AI system is giving over to you mark thanks Jim yeah uh we we discussed a lot about about risk management and it I'm sure a few of those bullet points there looks pretty familiar I mean a risk management process in general um is very similar in terms of first defining exactly what you're performing a risk a risk assessment against you need to determine what what is the business use case of the service and Define that specifically and then bring in your stakeholders right who's involved in developing Who's involved in using these services and then and

then kind of bring all of that together in terms of then identifying your risks and those risk categories are very similar whether we're talking about AI or if we're talking about General system services that that you still need to perform risk assessments against so that's where like we're kind of bringing in together ISO 420001 Miss AI RMF eui act but then also there's like Miss best practices around how to perform a risk assessment in general um the quality ISO 90001 a lot of these Frameworks have a lot of just general best practice in terms of Performing and managing a risk assessment that you could use for your AI systems as well um but now there

are a few nuances when it comes to Ai and Jim covered a couple of those in terms of the threat modeling and making sure that you know the the results accurate and fair and things like that um we'll cover them in a little bit more detail now as we talk about continuous monitoring um so firstly is like and Jim kind of mentioned this is you have to establish metrics that Define what is considered trustworthy and accurate within your AI system a couple of those metrics could be like what are what is considered correct predictions versus predictions that are considered incorrect there are also some other that you can identify like um making sure that the

the latency of the results is sufficient that the AI systems are are sufficiently up and running when you need them to be and then also like what exactly are the explanations that are being generated and if there are any sort of unexpected inputs and outputs that are happening within this AI system establishing those first metrics that you can continuously test throughout time it is the first step to really continuous monitoring and then next is implementing that real-time monitoring making sure that you have logging enabled making sure you have a way to analyze those inputs outputs that decisionmaking to determine when there is an anomaly and then determine okay what caused this anomaly how frequently is

this happening is there any way we can you know make any changes here to to improve the accuracy and fairet and the next is reviewing like any sort of reported incidents or errors that occur within the AI system monitoring how frequently these errors are incurring what are the similarities of these errors to kind of you know make sure that you're mitigating any sort of um you know issues that are happening with the AI system and then implementing a dashboard that allows logging and monitoring to be a lot more to be facilitated a lot easier will really increase the the chances of success in regards to your continuous monitoring system uh and then monitoring feedback directly

from users and Developers is a great way to get uh a lot of that feedback that you need to accurately monitor these systems over time whether that's introducing a feedback mechanism to allow like users to submit when there's bugs errors or even if they just run into s like an area where they could make a suggestion for the AI system or if they receive any sort of unexpected responses uh having that feedback me mechanism is a great way for you to kind of track all of those and then measure um what exactly is happening and then make any sort of changes that you need in order to successfully introduce changes to those systems um and

then another way to to gather feedback in regards to the AI systems that uses from the community implementing a process for reviewing reports in regards to you know what are the latest AI risks within from like thought leaders or through social forums and webinars um is a good way to keep on top of the technology and any sort of changes that these generative AI organizations are making and then taking all of these different feedback mechanisms whether it's the logging from the dashboards the user feedback mechanism where they're submitting bug reports and then also through like your research um having a way to manage all that feedback through a centralized Mech mechanism um is a good

way for you to continuously monitor those those CH changes the metrics and everything like that throughout the year uh so so yeah continuous monitoring is is probably uh one of the strongest and most important aspects of the risk management Jim anything you want to cover there I was yeah I was just going to say I would add on the feedback part I I think it's especially important with AI systems more than other systems you know even if you're going for ISO 27001 or something like that it specifically in regards to the fairness and bias concern in other words if you read the standards the one of the biggest mitigating controls that you can put in

place in order to assure that your system is unbiased and fair is to have open and kind of continuously monitored feedback mechanisms because it is it is the people using the AI systems that are going to be the on the front lines kind of the first to detect or suspect that you know uh recommendations may be incorrect or unfair or biased or whatever and so there's a lot of stress on you know effective feedback mechanisms in these standards for that reason because they're trying to avoid that bias and unfairness risk M it makes sense Jim it's hard to tune an automated system perfectly to make sure you're getting all the feedback you need so having

some sort of mechanism for the actual users to report when they see things or have suggestions it is a good way to get a lot of that feedback that you need cool so the next uh next section is in relation to using standards and and different Frameworks to help Implement not only a strong AI system but we'll also be talking about um some sort of different cyber security and privacy Frameworks as well so here's where we kind of dive all of the different AI Frameworks that are out there in the market uh I'm not going to go too deep but just kind of give you a quick overview of of what these different Frameworks are

uh how they kind of differentiate and what you need to know in order to you know determine which framework you you should really be uh looking at first or or working into your your specific systems so first we'll talk about the nist AI RMF uh which is which is built by nist and is available open source uh there are four to core functions of the risk AI RMF uh the first function is the Govern function which is generally what are the policies processes and procedures that you have in place related to the use of AI and and the AI risks and then also underg govern is the accountability of that AI use so making sure

that your users and and developers are uh you know Meeting those those policies and processes that you put in place and there's a accountability against them next is map which is where you establish the context of the use of AI categorization and also determining the Baseline and benchmarks for the performance of that AI measure monitoring and evaluating those AI systems Gathering feedback on the performance and then manage which is where you respond to prioritize manage risks risk treatment and Communications planning on how to handle incidents and negative impacts in regards to the risks also falls into that category as well n AI RMF uh also kind of uh overlaps with a lot of General Security

best practices such as uh incident response risk management as we talked about plenty and then also acceptable use and and specific processes related around the use of AI systems uh this framework is voluntarily is voluntary and also flexible so uh Al since it's not regulatory or not a law you don't need to meet this it's also pretty Flex where you don't need to meet it for T from top to bottom but basically implement it in a way that that best suit to your organization next is ISO 4201 which is also a framework designed for responsible development and use of AI systems it revolves around an aims which is which could be an extension of your

existing isms program which is for ISO 27,000 And1 so if you're already meeting 27,1 or that's something you're interested in for 21 is a great tack on where you can kind of implement all of those best practices that you have for 42 for 27,1 into your AI systems as well uh the the key thing about 4201 is it's also voluntary not a requirement but it can be attested against so you can have a third party perform an external audit and then you'll receive a report and that can tell you how well or if you need to make any improvements in regards to your control RS around AI next and last is the EU AI act

this is mandatory for organizations that are introducing AI into the European market the first set of regulations will be released next year and that starts with the the pro the prohibitions of using AI that is considered unacceptable risk all of those definitions are incredibly important unacceptable risks I'll just go through a few examples un acceptable would be something like social scoring or some forms of biometric identification and the use of AI within there there's high risk which could impact the health safety fundamental rights of people or the environment uh also has the highest level of restrictions um in regards to like the use of AI for educational critical infrastructure employment things like that the the

other two categories are limited risk and minimal risk which uh kind of fall into things like chat Bots or if you're using AI to do things like spam filters or game where the risk to the to public or Society or something like that is much much less limited we can jump into the next slide now we're just going to cover some general privacy and cyber security Frameworks these are important because as we went through that graphic you notice that a lot of fundamental controls are also important for AI so taking a look at some of these Frameworks could also help in regards to your use of of AI in a very effective and secure manner

firstly we'll talk about the Privacy Frameworks uh gdpr CCPA are two examples what they do is they mandate privacy controls for your organizations that are handling personal data if you're storing data in specific regions these uh Frameworks could be a requirement or law for you so it's important that you do review um you know what region you're storing personal data from making sure you're you're maintaining these these Frameworks uh and then next is ISO 270001 which we covered a little bit it's in regards to an isms uh maintains controls about uh data security it works really well with 420001 there's a lot of overlap there and then there's multiple different industry Frameworks Hippa PCI these

are all very specific to certain industries uh there's also Frameworks that you can look into and maintain if you're not in any particular you know industry that requires a cyber security framework those are things like SOC 2 ISO 270001 CIS or uh n 853 we can we can jump into the next section now so hopefully you're not overwhelmed at this point and you're really asking your question okay so where do I start um be because it's not all that complicated once you take a deep breath and step back a bit so um good segue from the standards that Mark just went over in that um you know the first step should be you know based

on where where you are geographically where your intended audience is geographically which of those standards or regulations uh you know make the most sense for you um in order to choose those Frameworks you probably need to familiarize yourself a little bit better the the nist uh framework is downloadable for free you do have to pay for the iso 420001 standard um you can get that from ansy in in the United States um or from ISO um you know the currency is Swiss Franks but it converts for you they're about I don't know $150 or something like that to download the standard so download the standard choose which framework you'd like to adhere to um understand

the requirements they're they're they're not that difficult to read actually if you you take your time they're actually pretty well put together you then need to decide okay how far off are you so the first thing you need to do is that Gap assessment you could even call it a risk assessment truthfully it's just getting your Baseline you know because you're already doing some things you're probably doing some good things probably not doing everything you need to do and so the important thing is well what's the extent of the gaps what's the nature of the gaps which of the gaps are most important to to close in what order and that's going to come from

the risk assessment part of it then then once you do you know when you start fixing these things especially uh if you want to get certified you know with an audit and you're going against ISO 420001 you need a way to organize all the evidence that you're actually doing these different things you're supposed to do that's where a platform like Secureframe comes in because it's so good at you know saying okay here's the requirements how you going to prove that where's the evidence download the evidence here and then it makes the auditor's job all that much easier yeah I'll jump into the first section that we have here which is in in regards to data

privacy and governance controls we'll be talking about cyber security controls a little bit later um but these are data privacy controls that are fundamental they can help you in order to maintain a a secure AI system we'll we'll cover a few of these here a lot of these are mentioned in um the gdpr or the CPR CCPA different uh privacy related Frameworks uh but first is in regards to how you ingest and collect personal data um making sure that you request consent or providing users a way to manage how their data is handled by your organization is usually a requirement that's typically through like site banners or terms of services that you have with your

users and through that terms of service you basically explain how that data is being used throughout its life cycle and usually also require some sort of Comm communication if those terms change after they've already been approved by by the users that sign that that terms of use um next is how do you manage this data securely first is determining who is managing that data to Define who your uh roles and responsibilities are in relation to ensuring the user data is ingested managed and disposed of properly and then also determining which data you're collecting is considered sensitive or if there are multiple categories classifying that data ensuring that sensitive user data is properly classified and differentiating

that data between your other forms of data is going to help you understand and make sure that the sensitive user data is only being stored processed transmitted within the specified environments that you have as secured enclaves for that data and then also minimizing the use of that data making sure that that collected sensitive personal data is only use when necessary and is not transmitted stored process for any unnecessary reason uh and then the retention of that data is that the last and and also important bit where you have a process for removing that data whether it's at a defined period of time or you have a workflow to have users request when that data should

be uh removed from your systems and then having a process for doing so at that point we'll jump into uh cyber security controls next yep thank you Mark now let's be clear we're listing three categories of cyber security controls here if you look at the center for Internet Security the CIS version 8.1 implementation groups there are 18 different control families so why did we choose these three we chose these three because these are the three that that in in our reading of the standards appear to be the the ones of highest priority um in terms of mitigating R risk for AI systems specifically so please don't misunderstand that if you take care of these three

you're all set you know if you go back to that diagram we did in the first place we expect you to have a comprehensive cyber security control program in place but as a relates to AI systems these are three of particular importance and I think you can see why in terms of Access Control you know you think about you know what would be the risk if you had lacks a access controls to your underlying data that your AI system depended on or to the rules engine that your AI system depended on so obviously you want to have very tight ACC access control again remember I was talking about the different roles that people have so

you want to have role-based access control only give people access to the to the um aspects different aspects of the system that they absolutely need to have in order to do their job multiactor authentication is an absolute must uh you know for AI systems for for all people really for you know from users on up it doesn't matter you in our opinion you should not deploy an AI system without strong multifactor authentication logging and monitoring again is not optional um you know if you look at the EU AI act that's legislation that's not a suggested standard it is super clear the level of monitoring the level of logging that you must have in place uh

as we said before if you don't know what's going on with your AI system how can you assure that it's making good recommendations in a fair and unbiased manner so it's sort of self-explanatory self-evident that logging and monitoring would have to be uh you know a critical control that have well established and then finally incident response and incident response in this case in the case of AI systems it's a little bit broader than just someone trying to break into your system you could look at you know even a report of of a biased finding as an incident right and so um you know especially in the EU with gdpr Etc they're very much about response

processes response to Consumers Etc and so this is incident response needs to take on a a broader definition when it comes to AI systems and it's not just that you know a bad guy broke into your system this is you responding to feedback from users as well great we'll jump into now why should you focus on security and and compliance in terms of your AI system and and Jim I'd love to hear your opinion on this too um to see if we have any uh different viewpoints on this but basically uh what I'm seeing in the market is you know as we're you know Secureframe as we reach out and and have calls with with

prospects they're asking us about not not just you know how the AI in our system works but how how is it implemented securely and I think this is turning into a a a big conversation now now that AI has been into the market for for a couple of years and it has really embedded itself into software and solutions the next question is you know it's not just how is AI being used in these systems it's how is it being used securely because now everybody is kind of understanding that there are risks related to the use of of AI um so so why should organizations go through this process I think uh the the first one

is just proving that you know the the customer's data or that prospect's data is a priority to you having a third-party organization audit your cyber security environment um and then generate a report stating that you are implementing these systems effectively uh is a great way to prove to your customers and your prospects that you're handling their data in a secure man and what that does is it helps you move through these deal Cycles faster where instead of having to explain how you may have some things in place being able to present a report that has the controls that you've implemented in relation to these systems that that you have established uh is a great way

to to answer a lot of your your prospects questions in relation to your cyber security best practices um and uh this is also now like I mentioned becoming a requirement it's it's not really just like a a nice have anymore a lot of you know prospects especially as you move up market like maybe startups or now it's like smaller businesses or medium-sized businesses are a lot more concerned about the the security of of the solutions that that you have in place as a third-party vendor or service provider to these organizations because your customers are have to maintain their own security as well and part of that is doing due diligence against you and making sure

that you have these these security controls in place um and then implementing AI securely is is becoming incredibly important uh it's on the the Forefront of all of our prospects questions and I'm sure you're going to be hearing it a lot as well now having a risk management program is going to put you ahead of that race you know a lot of these organizations are have just implemented Ai and they're just reading about these Frameworks and um if you're able to kind of kick off that that risk management process and that go start going down into M AI RMF ISO 4201 that's going to get you a head start and then once you get attested

against 4201 that's going to put you in the lead in in terms of some of your competitors who might not have that same uh attestation stating that you they're handling customer data as securely and effectively as you are and then like I kind of mentioned competing up Market you know as you as you talk to these bigger and bigger companies they're going to be asking you for specialized security Frameworks to CH to prove that you are implementing these controls effectively or they could be sending you um security questionnaires where you have to go through a list of control questions and explain how you have these things implemented there are ways to kind of potentially even

bypass these security questionaires by going through these Audits and being able to prove that you know you've gone through a rigorous AIT AIT it covers all the controls in these questionnaires and that can help you stand out and even you know get to the lead in terms of some of these deals that you're working through so using basically um security can be your competitive Advantage when it comes to these deals and sales Cycles especially when we're talking about technology that's so new as AI being on the Forefront of AI security can really help you stand out there Mark I'm glad you emphasized supply chain risk management and third-party risk management in regards to these systems

because that that's really critical you know that's become more of an emphasis for instance with the Department of Defense lately uh you know when they talk about a software bill of materials as a requirement for any systems developed for them but what's interesting in that regard is the majority of the our customers that we're helping with either ISO 420001 or the nist AI risk management framework are not what I would call end deployers of AI systems but they are companies that are developing AI systems for their customers so in other words they do the AI systems development process according to the standard and it's almost like they want their customers want like that Underwriters Laboratory

seal the equivalent of the Underwriters Laboratory seal on this product that comes out of our customer system before it gets delivered to their customer so you know I I think that Supply Chain management is very appropo to what we're talking about here M makes a lot of sense to me Jim I think like one of the first things you need to do is determine where you sit in terms of this AI landscape like if you're you know generating a lot of the uh pooled you know like kind of data for these AIS if you're just deploying aspects of AI within your system or if you're just a general user of a gen AI product you're

still you know you sit in this uh supply chain in terms of somewhere yeah exactly like and and you know how you use these vendors is going to be incredibly important when you when you try and move through uh ISO 420001 or any of these AI Frameworks don't make assumptions about the systems that you're using yeah exactly okay I think we got one more slide yep um so we've tried very hard to make this an objective and informative presentation for you keep the commercial part out of it and so we just wanted to indulge you uh with one slide a with a little bit more about Secureframe and Trava uh we are uh uh partners

with each other we work together in a very collaborative fashion you know it's um from a customers perspective it's it's almost like an integrated service if you will um you know what what uh Trava does is we are the compliance advisory well compliance is a Service Experts but we need a platform on which to do that that platform is Secureframe so Mark tell us a little bit about how that works from your perspective yeah we we're a platform that helps you organize all of the evidence the controls we even integrate within your your vendor technology stack to ensure that configurations are are set properly in relation to a lot of these compliance Frameworks so you

know what that means in layman's terms is you may be an organization and you don't have a compliance expert on your staff maybe you don't even have a cyber security person on staff you could be the CEO and founder that's trying to figure all of this out Secureframe we turn those complicated convolute Ed framework requirements into tasks b-size tests that are all in relation to the technology you integrate the Personnel that you invite to our platform and make make it an intuitive process for you to be able to go through these Frameworks step by step and then you have your compliant advisor Trava to be able to make sure that you're prioritizing these Tas effectively

to make sure that you're implementing these uh configuration changes effectively and you're doing things in the most uh effective manner possible to the point where you're going to go to audit and you're going to be fully prepared for some of these assessments so I know we talked in detail about a lot of these Frameworks and it may seem pretty complicated which it can be if you go and just download the the regulatory material but um you know with our help we've we've gotten many customers through these audits who might not have that that expertise on staff yeah Secureframe plus Trava is really the Easy Button for you in regards to this you know just one

more example if you wanted to be certified to the iso 420001 standard before the external audit will start their external audit you have to turn over your internal audit report most of our customers are too small to have an internal audit function so Trava does that internal audit for you just one example of how we we collectively Secureframe plus Trava get you absolutely ready for that on it and you will pass all right so that brings us to the end of today's presentation if you have any additional questions please feel free to ask you can use these QR codes to reach out to Trava or Secureframe but Mark and Jim before I let you to

go I know we covered a lot of information but is there something else you would like to talk about or really make sure that the viewers understand before heading out today Jim yeah I would say I don't want to sound like a broken record but I would go back to my do not pass go do not collect $200 if you don't understand the the source and the quality and the quality management of your data you're not ready to talk about building an AI system yeah my my thing is you might already have built that AI system and now we're just trying to uh go in reverse and try and make sure that we've done it

correctly oh my God what did we do exact exactly um honestly if if you don't know where to begin feel free to reach out to to Jim and myself like we we'll be more than happy to jump on a call um answer some emails in relation to any questions that you might have and really get you started um that might be the hardest part it it could seem like this giant effort and you know you might already have ai in place and it's like okay how do by reverse and go back so um if you do have questions about where to begin feel free to reach out to both of us um my QR code

is right here my email is marcrame decom feel free to shoot me an email if you do have any questions all right Mark and Jim thanks so much for your knowledge I hope everyone has a great day thanks jar bye