---
source: https://www.youtube.com/watch?v=mTqgsgsxeq0
title: ISO 42001 vs ISO 27001: The AI Governance Layer Your Program Is Missing
published: 2026-08-27
type: transcript (YouTube auto-captions, lightly corrected)
---

# ISO 42001 vs ISO 27001: The AI Governance Layer Your Program Is Missing

All right. Hello everyone and welcome. We are so glad you could join us for today's webinar. My name is Matt and I will be the host for today's session. Uh we'll be getting started in just a couple of minutes to give everyone a chance to log in. While we wait, I do have a few quick housekeeping notes to share. So first this webinar is being recorded and we will share the recording along with any additional resources after the session. So keep an eye on your inbox and we also encourage you to engage with us throughout the session. Um if you have any questions, drop them in the chat using the Q&A feature at any time.

Um we're going to do our best at the end to answer as many qu questions as possible uh during during the dedicated Q&A portion. So again, we'll give everyone a few minutes to join and then we will get started. We have a really exciting session today. Um hopefully you all find it valuable and we will get started shortly. If you're just joining us, grab a coffee or tea, settle in, and get ready for an insightful discussion. Uh we have great session ahead, and we're excited to dive in. We're going to give it about like one or two more minutes, and then we'll go ahead and get started. All right, I see a few more people

joining. Let's go ahead and get started. Um so today's webinar ISO 4201 the AI layer your ISO 2701 program is missing. I have two great speakers today. Um Caitlyn and Mario. Very excited to have you guys here and uh Mario care to introduce yourself? >> Absolutely. Thank you Matt. Uh welcome everyone. My name is Mario. Uh I'm the service lead for uh ISO at Insight Assurance and I'm also a a newly minted partner at the firm. Um we have worked with 42,0001's uh uh audits for about a year now. We've been accredited we were among the the first few accredititation bodies uh to become accredited to offer those services. Um and we are always happy

uh to share our knowledge uh and and insight no pun intended into the world of uh 4201. So I hope everybody enjoys. Thank you, Mario. We're also joined by Caitlyn today from Trava Security. Caitlyn, please introduce yourself. >> Hi, everyone. I'm Caitlyn Senoni. Great to be here. Um, I've worked with Trava Security as a security adviser walking uh organizations through ISO audits uh including 4201 um readiness and just operational and tactical levels. So, that's what I will be bringing to the table as well today. Awesome. Thank you for the introductions. So, let's go over what we're going to cover today. We're going to do an overview of uh the session agenda, ISO 4201, how it

relates to ISO 2001, what's driving demand for AI governance, and the path to certification. And then we're going to set some expectations. This is a practical conversation designed for organizations that already take security seriously and want to understand where AI governance fits. So getting started, what is ISO 4201? Mario, I will let you take it from here. >> Wonderful. So as with any ISO standards, uh it is meant to standardize a certain topic. In this particular case, we're dealing with um AI management systems. Um what that means in plain language is uh it is a standard that provides a scaffolding or the overall skeleton on how to govern the specific topic of AI within your

organization. Um it depends on what it is you want to achieve by deploying AI into your organization. But the spirit of 40201 is that it provides the structure and the necessary tools so that you can consistently achieve those objectives. The typical objectives that an organization u that we audit typically said are things like well we want to you use AI responsibly. We want to make sure that we're properly giving our risks the necessary coverage and treatment. uh and we want to make sure that the output of our AI systems or however it is that we're deploying um these new amazing tools um are okay for transparency, accountability as well as ethical considerations. Um so it

is uh unlike other frameworks that only focus on AI product development, ISO 4201 provides organizationwide governance on matters related to AI. Uh so it is not only limited uh to organizations that produce or develop or deploy AI but it also applies to really any organization that uses AI in any capacity under day-to-day operations. Uh so it could be argued that it is uh very widely applicable even perhaps more than 27,01 to organizations out there. Uh because many of us or really most of our organizations uh we have adopted AI in any way shape or form uh even if we're just using it or consuming it internally in order to aid or enhance our processes. I'll

give you a a very brief example. organizations that are using uh LLMs uh to assist on their day-to-day uh operations or perhaps even organizations that are um using AI to augment their staffing capacity or capabilities even if you're not developing that AI or you're not necessarily providing services to third parties using that AI. The moment you start consuming AI internally makes you allegible uh for 401. Um, so with that, uh, Caitlyn, what kind of organizations have you, uh, seen in in your time implementing that are typically, uh, moving forward with this particular standard? >> Yeah, great question. So I I will see a lot of organizations who don't really think of themselves as an AI

company necessarily but who essentially discovered that they use AI services um vendors uh processes their employees are using AI and they need more visibility and more management and governance there. Um, so even that, even if you're not building the AI models yourselves, even if you're not necessarily even producing a a product that focuses on an AI source, um, you might really want to think about 4201 as out how it applies to you. um even if even if you're scoping down to you know just an acceptable use policy AI inventory and just discovering you know how much your organization uses AI in general um can be really useful because it's definitely like Mario was saying it's

definitely not just for companies who produce AI or model AI for sure um yeah I know I've talked with a lot of organizations on the large side um that are just really worried about their employees especially just kind of using AI in their day-to-day copying in data that potentially you might not want in you know a public AI engine and how that looks for their organization. So >> yeah, absolutely. I um it's the same case with us. We see many organizations that are not necessarily again building or deploying AI. Uh but let's be let's be honest, most organizations even if they don't necessarily want to acknowledge it, um have employees or personnel that are consuming

AI. And generally turning a blind eye to that fact makes it so that u it doesn't make the the AI component go away. It just makes it so that it is now unmanaged risk. Um so even for uh people who typically feel like their organization is not um eligible for 42,0001 the answer is uh typically that yes there are individuals within your organization that are consuming AI they're using it whether you know it or not and ignoring that fact is not uh a way to actively manage those risks. So uh we've seen a very interesting mix of clients in in that arena. >> Yeah. I'll also just mention to the fact that um vendor vendor

and supplier risk there. A lot of times there can there can honestly just be a gap in knowledge um because a vendor that you might use might have pushed out an update that uses AI way more than than previously and if you know that might not be something [clears throat] that they push out to the rest of their customer base. So, if you're a part of their customer base, um you might be unknowingly using AI for a lot of your development or um customer success or or anything like that. And we'll get into that in the next couple slides as well, but just wanted to call that out that there can definitely be AI blind

spots. And so, just starting this conversation is is really important. >> Absolutely. >> Yeah. And with that, so getting into just the AI governance landscape today, um really there's essentially two areas to be aware of that I'm sure uh you all on this call have have had experience with at this point. Um one is that customer expectations are exponentially growing right now for AI security and governance um and visibility. And then there's a lot of regulatory uh developments as well both on the legal side um and on the framework side. So ISO 20 uh 4201 is really uh it's not a legal requirement. It's a it's a framework to build your management system on. Um

there are however several legal requirements that have become applicable in the last 2 years or so and that are growing uh and changing um in the next couple years that we can expect. So there's the EU AI act um specifically in forced since February 2025 I believe um that focuses on prohibited practices with use of AI AI literacy requirements things like that. There's uh a couple other just provider obligations, transparency obligations, um high-risk requirements uh that have taken more of a national aspect for enforcement. And then there's a lot of just uh various statebystate requirements as well that it's good to be aware of and just see if if any of them apply to you.

There's um California, Texas, New York um and several others that have uh essentially established uh those jurisdictional laws and requirements that kind of narrow it narrow it down a little bit. Um so definitely important to to have that scoping discussion define where your data is coming from where it's going to what processes and locations your uh AI use can touch um and can affect for sure because that that might really bring in you know different requirements or different legal uh legal actions that you need to take. Um, yeah. So, just in general, also from the customer side of things, I'm I'm sure we're all experienced with those amazing security questionnaires that that we all know

and love. Um, I've definitely been seeing a lot more AI governance security related questions there. Um, and I want to ask Mario just from your perspective there. Um, have you seen for for organizations that have come to you guys uh for 4201 auditing, have you seen more of a push from laws versus customer contracts right now? Is that is that you know one versus the other or is it kind of spread out in general? >> It's it's it's both, but at this point in time it's mostly uh contractual requirements from customers. And um I think the the the reason is uh as as all of you know 420,0001 is still a very recent standard. It

is something that uh only came to be a couple of years ago whereas some of more other established ISO standards have been around for decades. Um and for the first time ever, the very large, very big cloud service providers and the very uh you know the the the biggest players in the tech industry, they have just recently obtained their certifications in 4201. And when you do that, you can't really have a weaker link on your supply chain because that introduces unacceptable risks uh to these organizations. And now that they're certified, essentially those requirements for their supply chain to also become certified have started to trickle down. Um, and that's where uh we see many of

our clients um essentially scrambling now because now they have deadlines from uh their their biggest clients to become certified before this date uh or risk losing their contract. Um and so that's that's where the biggest um inflow of business uh has come for us recently. That being said, um AI regulation is still nassent. It is still a very in its infancy. Um and u the Europe which is always at the forefront of regulation, they have come up with the most comprehensive um set of regulations for AI and that has only just recently come into effect as well. uh meaning that we will also start seeing a big push of organizations that are going to try

to become 42,0001 certified um in order to have the tools necessary to fulfill these new uh regulatory obligations. And so right now we've seen mostly uh the inflow from client requirements but I'm sure that we will start seeing a huge inflow uh based on regulatory requirements as well very soon. Um and what I always uh tell our prospects or or clients is um listen this is something that is going to become a table stakes. This is something that is going to be required just to put your foot in the door uh to gain business especially if you are an AIdriven company. Uh so do it do it on your own terms. Don't wait for somebody

for a regulator or for a huge customer to put your company at risk or or maybe put a big deal at risk because you didn't do it uh early on and now you're scrambling uh to get this certification and you're not able to do it on your own terms basically how you would have loved to implement and deploy this management system in order to get real value out of it. Uh so uh I mean most of the individuals in this call uh they're likely uh probably in those shoes right now, right? and and they're the ones that have um made the determination that this is something that will be relevant for their company or they're

perhaps in the process of performing that implementation. Uh but I would always love to get this message out there uh to more uh information security and artificial intelligence and compliance professionals because we often see people that are caught backfooted u uh with these new waves of requirements. I was also going to just uh call out here just in general these are this is a kind of an everchanging landscape right um there's there's different uh proposals in progress right now there's some uh you know agreements on the EU side um that are that are in the proposal phase it's not really a law today um so it is good to just focus right now on what

is enforcable today for sure but it's also really important to keep uh keep in mind what could be enforceable a year from now and like Mario was saying to get started on that now um because you you never know you know how soon customers are going to start coming to you with this because it's come across their plate in their they're concerned that they want to make sure that their um their third party management is in place and and their security uh risk is in place as well. >> Absolutely. >> Oops. All righty. So, let's let's talk about uh 27,0001 versus 420,0001. Um I would love to know in the chat if anybody is familiar

or they already have 27,0001 implemented in their organization. Um because if that is the case, you will find that they are incredibly similar. Um ultimately there's a big misconception out there that ISO standards are about controls. Not really. ISO standards are frameworks for governance and and management, right? So when it comes to managing uh a certain topic, they're really the same principles essentially apply. So whether it comes to managing information security with 27,01 or artificial intelligence with 42,01 they are essentially the same principles and you will see that clearly reflected if you purchase a copy of each standard. Uh you will notice that the clauses um they are almost identical. essentially the body of requirements

of what you need to implement as an organization. The requirements that you need to keep in mind, they're going to be almost identical except one is dealing with um a focus on information security and the other a focus on artificial intelligence. So um you know things like uh the differences start where uh there is topic specific controls or risks that need to be addressed. uh in the case of AI it's mostly around the responsible development deployment and oversight of those AI systems as well as the the the provenence of the data the guardrails that you put in put in place and the ethical considerations that you need to keep in mind that being said that

we're talking about the control set specifically where most of the differences come in when it comes to the actual meat and potatoes of the standard the clauses four through 10 which contain the most requirements they're going to be almost identical which means that it is fairly easy to layer one on top of the other because the bones the skeleton are very similar and we'll dive a little bit more into that concept later. Um so these two standards are complimentary and not competing. Uh meaning that uh they are able to build into each other because of course artificial intelligence systems are also information systems where you need to make sure that information security is in place.

uh if not like any other information system uh it is bound to uh essentially uh you know get breached uh to filter information that perhaps shouldn't have been out there into the world. Uh but more seriously it is bound to potentially make decisions based on data that wasn't the correct data. Um, and just like an employee, you wouldn't have an employee in your company that is um that is incompetent or an employee that is not giving you the necessary results um or an employee that is not trustworthy. And so you need all of these elements to be in place in order for your AI systems to be trustworthy both internally and externally. And that means

that you need the right data, you need the right information, and you need to secure it properly. So, uh it's not one or the other. They can both coexist. And in fact, they are designed to coexist and live together. >> Yeah, I would 100% agree, Maro. The from the from the audit standpoint, just just in general, there's, you know, you could you could do an integrated audit for those of you who already have uh ISO 27001 um and are in your surveillance uh your surveillance time frame. essentially adding 4201 uh is not going to be that big of a leap from starting from scratch. Certainly um there's definitely areas that 421 essentially assumes you already

have uh from the security management perspective like an incident response process, change management, um a risk assessment methodology, right? Um so those are very basic kind of baseline uh information security processes that are required for an information management system. Um and 4201 just adds on like Mario was saying that that AI perspective um and that AI security and governance aspect. Um, so you could potentially very seamlessly I'm sure do a do a integrated audit um adding on uh so that you know and and Mario I'd love to hear perspective of kind of how this looks from the auditor side too of of just the testing and kind of how that looks um in general when

you know an organization wants to do both at the same time um because again it's not separate controls necessarily for a lot of it. Yeah, that's uh that's where a skillful auditor is going to be necessary because ultimately most of the requirements are are very similar because we're talking again about governance uh and governance is governance whether you're governing AI or information security but when it comes to the technical controls and and where those differences in both standards are um are found um you need a skillful auditor to be able to trace a single piece of evidence that can be applied to more than one requirement and you need them to be uh wellversed enough

on both standards so that they can understand and map it appropriately without you having to answer the same question uh two or three times right uh so it is um um the the key in order to do that is one you need to have a a well-trained and experienced auditor but you also uh need to make sure that you perform scoping correctly with the client that you really understand their system their context because that is um It's just going to vary substantially from organization to organization and a skillful auditor needs to not check the box because anybody can take a piece of evidence and say it looks good right uh but it is within the

organizational context of the client um are we materially hitting the objectives that the client has set out for themselves. So it's essentially a a mix of uh you need a really skillful and good auditor that is able to uh really really understand and tailor the audit program uh to your specific organization. >> Definitely. >> Let's >> All right. So yeah, jumping into essentially how those two standards intersect, right? So there's a lot of areas that overlap. I want to just kind of call them out specifically for you all today. Um essentially the risk management again it essentially assumes you already have a riskmanagement process risk management methodology. Um but for uh for 4201 it's going

to be essentially the same process. Identify, assess, treat. Um what changes is the risk categories that you're going to be scoring. So um basically you're going to be looking at making sure a system doesn't produce a biased outcome. um making sure you can explain why your AI system made a call one way or the other. Um what's the logic behind it uh making sure that you understand what happens if um an action is taken by AI before there's a chance for human uh review, right? Um so that's all going to be included in the AI portion of risk management. And then supplier controls is a really big one. Um, so again using the same basic

vendor due diligence process that you already have but including more of an AI data provider perspective. Um, so not just necessarily asking for a vendor's SOC 2 report. Um, but also again looking to see if if your vendor potentially, you know, added some some kind of AI feature last year and they didn't tell you about it. You know, what are they actually doing? where did the training data come from? Has anyone checked for bias? All of that um is going to need to be a part of that supplier control management. Um the incident management process is another one. Um so a lot of incident management on the information security side uh has to do with

um you know data leaks, insider threats, um you know ransomware, that type of thing. But from an AI perspective, uh, your incidents are going to look a little bit different. It's going to be more something something like, uh, what we talked about for risk management, right? It's going to be did, uh, a biased outcome occur, right? Was there, uh, you know, a hiring tool that somehow has rules set that automatically, um, automatically discards, you know, applicants above a certain age. um or uh you know was was a chatbot telling a customer something that was completely faulty um because you know in that case you know now that organization is accountable for what the chatbot say

said. So um there's there's been organizations who've had to pay you know hundreds of thousands of dollars to their customers because the chatbot said they would. [snorts] Um and that can definitely be a really big risk. um definitely going to need to be followed your following your incident response process there as well. Um and then governance uh is another really big area there where where there's intersection. Um we've already talked about you know just in general um information management and AI management are very very very similar when it comes to governance. uh your policy review cadence, leadership commitment, roles and responsibilities, all of that um essentially can just get broadened to include the AI aspect

as well without having to build a whole new structure there. Um yeah, so there is a couple areas where 27001 does stop helping. Um so 4201 has a couple specific AI things like AI impact assessments um data training uh and governance there um management in your software development life cycle [clears throat] if AI is a part of that and then uh also as we talked about AI specific vendor criteria as well. So that that's really important areas to keep in mind when it comes to uh if you're starting with 27001 as your backbone. Um very similar to SOC 2 as well. If you're starting with that um very it's that's very similar to ISO 27001

as well. So it's uh that kind of carries over there there just about as well. So >> Mario, I wanted to ask uh just from your perspective um with 4201 um have you seen just in general like the areas that that I kind of just mentioned um if there's any that cause more friction in an audit like things that you know people forget about or just didn't build out properly. Oh, Mario, I think you're muted. >> I appreciate it. Yes, I was going to say the most common pitfall has to do with the AI impact assessment precisely because it is one of the few areas where 420,0001 differs from 27,01. And if you're not paying

attention during the implementation, uh you may just miss it. uh which is a a big issue because uh a lot of the actions that you take um in terms of your 420,0001 implementation uh are are driven by the results of your AI impact assessment. Namely, how uh your use of AI systems uh impacts um your organization, your stakeholders, society in general, your clients, essentially understanding the impact of the systems that you're going to put out into the world. So um many organizations that are coming from 27,01 um they basically expand their existing policies u to also include 420,0001 elements but then they forget one of the more key components of it which is that AI

impact assessment. That's probably where we see the most uh uh things uh in the audit. Righty let's move on uh let's talk about what 4201 actually requires. So uh there is uh again this is a big area of confusion and and a lot of there are a lot of misconceptions around this topic. Um 4201 is a a wide open canvas like you can make it into anything you want to make it. It's a little bit of a choose your own adventure um kind of a deal where as an organization you need to determine what success looks like for your organization and then you set objectives based on on on that business case because ultimately you're

doing this not necessarily well maybe because somebody's requiring you to do so but also because like anything else in business you are uh probably trying to hit uh certain KPIs you're trying to be data driven you're trying to move u towards the strategic direction that executive management has set and that requires you to establish what success looks like through clear and measurable objectives. Um, and then everything revolves around that including things like uh risk management and the controls that you select. Those are all in the service of you hitting your objectives as an organization. Why do we manage risk? Uh because those risks are risks to hitting our objectives, right? If we want to achieve

a certain outcome or reach a a certain intended result, then we need to make sure that we're dealing with the risks that we might not actually reach those results. So when you contextualize um the standard in that way, it it it makes it fairly plain to see that this is just another business tool for the organization and not necessarily something that you're doing solely because you need to comply or you need to have the certificate for a big client of yours, right? And so um things like the uh once you set those objectives and you need to perform that risk assessment, you need to make sure that you based on that risk assessment select the

appropriate controls that you are going to be implementing. Um and things like how you manage your incidents uh and other organizational governance components. uh as well as understanding where um 40 thou 42,0001 objectives may align with 27,0001 objectives because they're ultimately part of the same management system just governing different topics. But you always need to remember that it is ultimately the same management team that is managing the organization. And so it's not like you have a management team for uh 27,0001 and another for 42,0001. it is ultimately the same team of people trying to move the organization towards a strategic goal that executive management has set. So um that is what 4201 requires. It basically

requires anything or any actions that you need to take in order to reach those organizational objectives right um and everything else is in service of that. So um just wanted to clear that that big misconception that it is just a certain set of controls that you need to implement or a like a set or fixed uh set of actions that you need to take. It's basically a you have to custom build it for your own organization based on what you are trying to achieve. Um this is typically one of the more difficult areas that that clients have just trying to conceptualize that uh and that is where a a strong uh consultancy or implementation assistance

might come in handy. U Caitlyn can you tell us a little bit more about how uh how what the journey looks like from an impletor uh or a consultancy firm perspective? >> Yes. Yeah. So definitely really important to remember that for all essentially ISO framework uh certification paths um like Mario said it's very risk based it's very management uh management needs to review the risk review the organization objectives and then the controls that are put in place on the technical uh and operational side um need to be built to meet those objectives um and any potential risks. Any potential controls that are not implemented um need to have very uh very defined very well- definfined

uh risk management uh or risk acceptance um from management essentially. So it's it's on you as the organization to determine um what risks you are going to mitigate, accept um or transfer etc. uh and what controls are going to be in place to uh to remediate those risks from a management perspective. Um so on the technical side, there's a lot of different controls. [clears throat] There's so many as I'm sure a lot of you know if you've gone through SOC 2 or ISO 27,0001. Um but some of the really big ones that I like to call out there would be uh you know just inventory tracking for for AI. Um, a lot of times, you

know, no matter what your scope is, that's going to be something that everyone needs to accomplish. Um, CI/CD in your in your development pipeline. Um, having some kind of uh blocking mechanism when when you have an AI AI in your development pipeline, making sure that you have impact and risk assessment uh blocks there. Um and then again just kind of depending on what you use it for. Um if you are actually developing models that's going to be a different set of controls right that you have to look at and review. And then if you're just using, you know, chat GPT or or claude or something like that, then that would be potentially uh another set,

you could you could maybe kind of scope down the recommended controls um and and review those for for risk in your management team. Yeah, it's definitely not a a technical spec. Um it's a a management system standard. So it's it's to be used to it's to be used by management and by your technical team uh to determine what your organization needs to implement based on the risk of your organization. And then moving on from that just to to go into a little more detail here about really the steps from getting from you know zero to hero for your certification right um so if you all have uh ISO 2701 you're very familiar with the process

for the stage 1 to the stage 2 audit um and all of the preparation before that for uh for AI for 421 it's going to be very very similar um highly recommended to Start out with a gap assessment uh to just see what your current state is compared to the standard. Um and then again build out that AI data inventory. See where your company is using AI from internal and then also from the supplier standpoint. Um complete that risk assessment and impact assessment coming from that AI perspective. And then the really really big one here is going to be the scoping decisions for your organization. Um so ISO 4200,0001 has a statement of applicability just

like ISO 27,0001 does. Um, however, it's going to really depend on whether or not you're an organization that um, essentially just uses an AI tool in your day-to-day, produces an AI tool, builds it yourself uh, and produces it for others or um, potentially if you use a tool that someone else built for you and you've built it into a product that you then give to another customer. So all of those different questions are really necessary to be asked. Definitely uh definitely helpful if you have some kind of advisory or or just uh someone to kind of discuss that with on the audit side as well um to figure out where you stand as an organization

because that decision will impact all the rest of the controls that potentially could be in scope for you. Um >> that is a great point. I'm sorry to interrupt, but I I have found that um organizations that seek consultancy or implementation help that are the most successful, >> the the key differentiator is that their consultant help them work through the right questions, right? um instead of prescribing uh like a an outofthe-box solution that is quote unquote widely applicable, it is the consultants that have the client walk through the important questions and the important scoping and really understanding the context that actually produce the the best results. The best results being the clients that uh actually

have management systems that operate as intended and that achieve the intended results. So, uh, that that having the the right person on your corner to help you navigate and ask the important questions instead of telling you what you need to do, but rather helping you navigate and reach the right conclusions so that you can build a tailor made management system. That is typically what we see on the audit side. Um, essentially shows the highest rate of success for organizations that have sought consultancy. >> For sure. Mario, have you seen organizations uh potentially even in like the stage one where you're where you're first reviewing their statement of applicability, have you seen them um scope themselves

higher uh or or broader uh than they should be and what would be the implications of that in an audit? >> Absolutely. Um, and the scoping is common. Um, and and and we have a process to try to catch it as early as possible because we don't want to lose our clients time if they're actually requesting an audit for which they are improperly scoped. That is part of our application review process and part of our planning process. But, uh, typically uh, and we see it both ways. We see clients over scoping or underscoping um, and there is um, you know, we can't certify something that isn't there. So if the client has uh mistakenly determined

their role and u and that's typically an issue with like out ofthe-box solutions that don't necessarily take into consideration the peculiarities of a given client um then they may have a bunch of policies and controls and a lot of things that do not reflect reality because they're not that particular role. And that's where the value of asking the right questions and taking the time to properly sculpt it from the beginning um really really pays off. Uh it really uh deres a big return on on that investment of time to actually sit down and ask the right questions from the beginning. >> Yeah. And speaking of that like next steps essentially once you determine your scope

would be uh policy creation. If you already have a policy set, you know, adding on an AI acceptable use policy does not necessarily have to be a completely separate document. You can add on to documents that you already have. Um, but you know, there's going to be specific AI scoping policies, roles, risk, impact procedures, um, data governance procedure, etc. that you need to make sure you're including the AI perspective with that as well. Um, and then just your internal audit, your management review, you know, all of those good ISO steps before your stage one audit. That's recommended before your stage one audit. And then um moving into that audit uh you know, reaching out to

certifying body and then uh Mario, what would be the the common, you know, time frame for uh the stage one audit, remediation window, stage two audit, certificate issued, etc. Well, this varies wildly from organization to organization, but um and it typically depends on organizational readiness. So, typically the the most time spent uh in the certification process is actually the client uh remediating things uh or or maybe uh them actually being ready to move on to the next stage. But if we assume a perfect world where the client is absolutely ready um and we we're not on a capacity crunch because that always happens closer to December. Everybody wants their certificates at the same time. Um

what typically happens is uh a client would sign an engagement letter with an accredited certification body like ourselves. Um and then we would within five days we would on board the client. We uh request um certain application information beforehand in order to one ensure that we are qualified to perform that audit for this particular client given their uh their scope and uh b that we have all of the information necessary to actually plan the audit, plan the engagement. So that happens typically within five days of on boarding. Then if everything looks good, we have the uh information we need, then we schedule the client for stage one as soon as u the client is ready

to undergo that stage one. That's typically a week or maybe uh between a week to three weeks out. Uh once that stage one takes place, we essentially know if the client is in fact ready to move on to stage two or if we need to provide feedback so that they're able to address any areas of concern before we actually approve them to move on to stage two. And then we already have a much much better picture of the client scope, their context and their management system so that we can actually very precisely plan out a stage two as efficiently as possible. Then in between stage one and stage two. Uh and that stage one typically

lasts about two days. But in between stage one and stage two, we recommend clients leave at least three weeks in between because if we do find areas of concern, we want to make sure that the client has at least a couple of weeks to try to address them and come up with a game plan on how they're going to fix it before that stage two. Um, however, organizations that are um, you know, if they're really ready to go, if they're really well prepared, they have no areas of concern, we can potentially move up the timeline a little bit. Uh, and then on the other side of the spectrum, we have clients that they're definitely not

ready. Perhaps they overestimated um, or underestimated how um, you know, easy the preparation would be and then we may need to push out the stage student indefinitely until they're able to reassess and regroup. Um however once we reach to uh the stage two then um how long that's going to take is going to be fully dependent on client characteristics u because there are accreditation rules that mandate how long we need to spend on that audit time with the client. So things like headcount, things like uh client complexity, the risk uh introduced by the their AI systems, the industry they're in, uh how many regulatory requirements are subject to, all of that adds to the audit

time. Um and a and a good accredited and serious certification body is going to take as much time as they need in order to ensure that the management system is in fact compliant and effective. Um, so, uh, it's typically a little bit of a red flag when you quote out with a certification body and a they're not accredited, but also b, um, they're they're telling you they can do this in in a couple of days, right? Typically, that's not the case. Uh, especially when it comes to the entire certification timeline. It is not unusual that organizations take anywhere from perhaps three months on the short end all the way to eight months up to a

year on the long end depending on again organizational complexity, size as well as um how well prepared they actually are for the the audit. Now um let's take a look at what organizations should be doing now. Right? If you are just about to start uh or embark on this journey, what what do you need to pay attention to? So the first thing is um you know start with an inventory. The first thing you need to understand is which AI systems you're using. Um and not only the ones that you're officially using but what are your people using right a little bit of shadow IT is always present in the mix and it is important that

you understand what your people are using so that you can actually govern it right. So maybe it is a tool that um you will not accept or it is a tool that you will accept but you will accept under your own conditions as an organization. So um it is important that you do not turn a blind eye to the tools that your team is using. So once you start with that inventory then you can start asking all the right questions regarding sculpting and context and objectives and then you move on to a risk assessment. You essentially approach the standard in order of clauses and that will give you a good guidance or when in doubt

because many of us here are are not ISO experts. Um then reach out to a qualified ISO expert or ISO professional that is going to be able to help you navigate through asking the right questions and through selecting the right scope and setting the right context and establishing the necessary controls. Um, so, uh, Caitlyn, how do you typically work with organizations, uh, where maybe they have not really implemented any management system in the past, but now they're seeking 42,0001 for the first time. How does a a company uh, and a professional like you help them navigate that process? >> That's a great question. I typically uh, really like to get them started just on the

scoping question. um determining what type of organization, you know, what their use case is for AI um in general and then that can often scope them down quite a bit on on the type of controls they need to focus on. But um I would definitely agree that the inventory step is a huge step. Um making sure you know everywhere AI uh is implemented within your organization. Um and then risk taring different use cases as well um before you get to documentation. [clears throat] Once you have that essentially um uh that inventory and risk framework to build on um you can start building out your your processes and procedures there. Um if you are starting with

27001 um you have a lot more to build on which is great. If you're starting from scratch then it's essentially uh you know I would recommend using some some framework similar to ISO 2701 to get you started there as well. >> All right. Thank you guys for that insightful session. Um, we're going to move on to key takeaways and Q&A. Won't go through all these because I know we have a lot of questions in the chat that that want that need to be answered. Uh, so you guys can just take a second here and and look at that slide. Um, but moving on to Q&A. We had a lot of questions. Um, so how this

works is we are going to go by the most up votes. So, if you want to see a question answered, feel free to vote on that in the Q&A section. Um, but our first question here is from George. So, do you find customers actually accepting 420,0001 as compliance evidence? Our org provides 27,0001 stock type 2 HIPPA on most infosc questionnaires, but still get asked redundant questions as part of our infosc reviews, and those are relatively old frameworks. George, I am sorry um to be the bearer of bad news. Uh we will never ever get rid of compliance questionnaires. Maybe in the future, but in any case, um it depends on the sophistication of the procurement

department that you're dealing with. Um many of them um that understand what the requirements are and how they're mapped, they're going to be able to leverage uh the certificate, but you need to bear in mind that the certificate does not really contain a lot of information. It's just a certificate outlining what the scope is and what was uh covered. but it doesn't necessarily provide or fully provide all of the answers that a procurement department might need. That's where you may need to provide the report. Uh in any case, I guarantee you that on average, if you're looking at the total amount of hours that your team spends filling out security questionnaires, you're going to see

a significant um u uh you know, move down if you do have a a 42,0001 ISO certificate. That being said, you will still get asked redundant questions. Um that's just part of procurement in general. >> Yeah, definitely. For sure. I I would say the uh the questionnaire aspect is unfortunately never going to go away. But um it's also helpful to let customers know if potentially you haven't reached that 4201 uh certification yet. um that potentially it's something you're working towards and you can always provide them your progress there of you know an AI acceptable use policy um training uh for personnel um you know checks and balances in your development pipeline that kind of thing

can also really be helpful. >> Awesome. Next question here from Mark is RPA robotic process automation topics part of the ISO 4201 specification. So um not explicitly there are no specific controls that specifically address RPA. Uh that being said remember this is uh anything you want it to be standard. So if part of your organizational objectives is to um it depends on what you want to achieve with RPA. Whether it is um that you want to achieve additional efficiency by using RPA, right? That's part of your AI objectives or whether it is you want to use use it securely or use RPA ethically, whatever it is you're trying to achieve with RPA. Once you enshrine

it amongst your 420,0001 objectives, then um that is essentially the entire goal of the 420,0001 standard. So there are controls that will be useful in addressing whatever it is your objective is regarding RPA. But there is a nice feature on the ISO standards and not a lot of people know about which is you can bring controls from any framework. You don't have to use the ISO controls if you don't want to. uh if your risk assessment results are that you are better off with another set of controls, you can fold that set of controls into your ISO implementation and we as auditors will certify you against that specific control set. So even though ISO 42,0001

doesn't explicitly address RPA, you can address it by making it part of your objectives and you can also bring in any controls that you deem necessary in order to achieve those objectives regarding RPA, even if they're not ISO control. So you can import your controls, you can make them up, you can design your own controls. Uh so yeah, doesn't explicitly address RPA, but it it it can if you wanted. >> And a big question to ask yourself when you're determining what systems or processes to include uh in scope for your audit would be, you know, what processes or or products are your customers going to ask you about, right? So if it essentially a good

rule of thumb is if it's uh is for the customer or touches production in any way those are uh those are important systems to include especially if it includes automation um especially if it includes uh you know autonomous automation as well. Um so if it's if it's AI making sure it's included in your procedures and processes and security management um for for those checks and balances. That is a great point, Caitlyn. >> All right, next question here. Which ISO4201 controls are the hardest to produce audit evidence for and what evidence do auditors commonly expect? >> I this is a a very difficult to answer question because it is really broad. Um the the truth uh

Cavala is it is um it's going to depend um very very u intensely on what it is that we're talking about what the scope is and what the the client is doing. That being said, um, typically anything that has to do with you demonstrating that whatever is going on within the the black box is, uh, working as for your policies and guard rails is going to be really difficult to evidence unless you have really robust documentation, um, around surrounding how you control uh, those outputs and and basically what um, that that the output of that processing is set in your own parameters. So this is again going to depend on your role. It's going to

depend on what you're doing with these systems. It's going to depend on um on on a million different factors. So it's a a fairly difficult question to to address appropriately. >> I'd definitely say from the tactical like day-to-day aspect of it. Um if it's a if it's a process that's done ad hoc, that's going to be very difficult to produce consistent evidence for as well. If it's a process that's repeatable in your policies and that your team is trained on to produce the same result multiple times uh or just continuously that is going to be a little bit more uh streamlined to produce evidence for to provide evidence for. Um and then if you use

some type of uh GRC uh audit you know evidence collection tool that can be very helpful as well to to essentially collect that. My advice is when in doubt, uh, document it. Uh, just in case you ever need it for the auditors. Right. >> All right. We have one more. We have time for one more question here. Um, so this one is another one from Mark. Is getting a ISO 2001 a prerequisite to getting a ISO4201 certification? We're on the journey to get to SOC 2 type 23. uh we know there are many SOC 2 controls that map to ISO 2001. We estimate based on our SOC 2 controls about 70% will directly map. Uh

the good news is no it is not a prerequisite to have 27,01 in order to get 42,0001. They can live together. They can coexist. And uh to also address Gabila's other question, we can even audit them jointly. We can perform a an integrated management system audit. Um but it is they're not they're not a requirement. If you want to have one and not the other uh you can absolutely do that. Um now when it comes to the control overlap um again between sock and 27,0001 you'll see different numbers quoted because uh it depends on the controls you select for 27,0001 as well as the um the the the control criteria or the um the audit

criteria for sock. But 70% sounds like a like a on the higher end but it sounds feasible. Um that being said there are um it it is a lower percentage for uh 42,0001 precisely because it is sock and 40 and and 27,0001 are covering the same topic of roughly information security protecting CIA whereas 4201 is dealing with a completely different topic. So the overlap is much lower. We typically put it at 30 to 40% on the highest end. And I also think it's it's good to call out that that for SOC 2 and ISO the controls the technical controls do have a very high overlap again just depending on your scoping. Um but it's it's

really good to remember that for any ISO framework um we there's also requirements which are a lot more uh documentation based um management reviews internal audits etc. um that SOC 2 doesn't necessarily require. Um so keeping that in mind that 70% you know might be uh a little bit high but from the control standpoint from the technical standpoint um I would say it it might be a lot more accurate but from the documentation standpoint uh for any ISO uh framework you're going to be adding more documentation uh ISMS documentation and management review as well. Um, so keeping that in mind from the technical side of things, it might not be as big of a lift.

From the documentation side of things and admin side of things, it might be a little bit more of a lift. >> Awesome. Well, thank you guys for a fantastic Q&A session. We tried to answer as many questions as we could, but if you do want to learn more about ISO 4201, we have another webinar coming up. So, if you guys would like to register for that, you can do that now on the U popup right there. And Mario, Caitlyn, I want to thank you guys for an amazing session that was very insightful and I hope all of our viewers enjoyed it. Um, so if you guys want to get in touch with Insight, uh, we

have our email there for sales insighturance.com as well as for Trava. You guys can reach them at salestravascurity.com. And um thank you again. Any last words, Mario and Caitlyn, before we sign off? >> I had a lot of fun. Thank you, and everybody who joined. >> Awesome. Well, thank you guys, and I hope everyone has a good rest of your day. Bye.
