AI in Risk Management – Special Interest Group launch Webinar

AI glossery we’ve produced is that right

yes that’s right um what the glossery

intends to address is to help organis

organizations recognize whether they’re

using Ai and manage the risks this is

not a technical publication it

essentially walks you through the basics

and when we’re putting this together uh

Andrew and I did work on this together

and we agreed 15 terms that we felt were

probably the most commonly used but to

focus on the top 15 uh and then to

provide a summary of the impacts across

a number of business functions and

domains from strategic Financial

reputation operational and legal

considerations and we have covered the

crucial definition of AI and um I picked

actually one of the definitions to give

you an example of what is in here and um

the pi picking the obvious one is the AI

system terminology because that’s the

most important and um what we’ve covered

is a high level definition so it’s

designed it’s a machine system computer

program designed to simulate human

intelligence but we’ve also included the

legal definition which has been adopted

in the EU and that’s originated from the

oecd AI principles which is a wealth of

resources if you are just learning about

ai go and have a look at oecd doai we’ve

also describe the use cases and in Risk

you’ve got fraud detection um there’s

medical diagnosis fa faal recognition

software but also to look at the key

risk considerations such as the

Strategic considerations the

overreliance on AI lack of transparency

maybe there’s not enough governance in

there or understanding of the technology

and finally the the purpose here is to

encourage the use of AI to uh augment

the work of humans uh to show that it

can work across functions that it can

work proactively and that it can be part

of a business product or service but

also to show it needs to be treated

differently to cyber security or data

privacy those are side effects of using

uh technology the key considerations

that need to be looked at if you’re

starting out and you know procuring

you’re designing in AI the glossery

should be your go-to place just to

explain some of those terms that you may

not have heard before so that’s that’s a

a potted overview and I would encourage

you know the audience to go and download

that and and have a look and certainly

provide

feedback fantastic thank you Pauline

and um Peter if I start with you if I

could ask you what do you see as the

risks and opportunities for AI in this

year yes thank you Andrew uh as we uh

could see from latest reports of using

AI in

organization uh it is not so popular AI

could be democratized through

application of uh

simple simple chart gpts

we have example here from Poland uh we

developed in our community polish llm

called bik it it

recognize country context and it could

be therefore used better for uh for

example government proposes because it

understands the issues what is a risk

yesterday polish Daily Newspaper one of

the biggest published a paper that over

90% of people share companies and

sensitive data with public llms that

means chart GPT or Cloud so thanks Peter

there’s a lot there around the idea that

one of the greatest risks is that we

lose our privacy and we share overshare

without knowing we’re oversharing um any

Reflections there on the on the risks

and opportunities and that that idea of

oversharing the opportunities which

everybody will will readily understand

are immense but the challenges for most

organ ganizations right now are

understanding in how to actually move

forward a number of organizations are

trying to go through and become AI ready

within that there’s everything from

understanding their digital footprint

and digital maturity understanding their

human capital as we all know in the

world of technology it’s easy to unplug

or plug in New pieces of Technology but

if the people themselves don’t actually

use it or don’t know how to use it or

refuse to use the new technology you

don’t get the benefit from it we see a a

lot of really interesting challenges

with organizations that say well I want

to use it for this but the technology

isn’t there yet or they don’t understand

their workflow well enough to we the llm

or whatever that the system is they’re

trying to use and then starting to think

through the data element what does that

look like the next big challenge which a

lot of organizations are being really

hesitant of is understanding what the

financials are within the AI space so if

I try and use some of these systems do I

get a financial benefit and then really

the last one is the risk in the

governance piece and that’s very very

much into roles and tasks at least for

us right now who has to do what within

your organization what committees need

to be formed is this a risk committee is

auto committee is it an AI committee do

you need a chief AI officer for example

to be on the line as as all of this and

then going through the organization to

make sure you have all the plans

policies and procedures in place to use

it those are some of the real sort of

significant challenges we see at the

moment but when you’re putting AI into

an

organization uh how do you stack up the

financial case and what do you see the

are the opportunities and risks for AI

That’s a super really good question

there Andrew I think we can maybe use

some of the previous experiences going

from maybe pen and paper into more

Technology based tools and say okay we

you know our risk management program is

now on paper and it’s in Excel

spreadsheets and now you want me to buy

this fancy tool it’s the same cons

considerations um but it’s not my

experience that buying better tools and

and increasing let’s say

efficiency um actually leads

organizations to say oh then we’re going

to like cut Effectiveness across our

organization and Let Go part of our team

members so are we talking efficiency I’m

going to reduce my team or Effectiveness

right I’m going to be spending my

existing time better removing man manual

tasks and getting more out of the team

have or maybe I have to build the team

in a different way with different

capabilities than today uh so I can root

out some of the more manual task that

would be my Approach and not say oh

you’re buying AI so you can save a lot

of money talking challenges or risks in

in AI ER seen from my perspective ER

being in the security and the physical

security realm here or industry the lack

of insight and knowledge from Risk

Managers typically is a huge barrier

that we’re not going to get the best of

this technology I can also see that some

of the organizations we work with do not

necessarily address the challenges that

the risk manager has but maybe like core

into financial and operations because

that’s where the money goes so risk

manager are some sometimes LIF a little

bit alone to navigate this space how can

we use it and if we identify a reason to

do so how do we get the money to to buy

the the skills because the existing AI

Engineers within the company they’re

busy optimizing the supply chain or

whatever so I think that’s a huge

challenge to be honest

opportunities very short in my space

where we talking physical security how

do we analyze vast amounts of

information uh sometimes it’s called

intelligence and translate that into to

action yeah have you got any examples

that you can think of from from your

experience of uh AI doing well doing

badly screw-ups that kind of thing all

the examples I’m going to give you are

actually based on classic AIS not even

gen AI just to underline the reason why

action was taken in Europe to create the

EU AI Act and the origins of the the the

law um partly originated from an issue

that arose in the Dutch government um

and this issue crystallized during the

pandemic in a nutshell they’d use

classic AI techniques to develop an

automated system uh Within child welfare

um department and that automated system

automatically sent out reclaims of child

welfare benefits to people who were of

dual nationality that singled out people

who weren’t Dutch Nationals and clawed

back all this money and um caused Untold

harm to these people and it prevailed

for about five years without it being

challenged fully and it was Amnesty

International actually that raised this

issue amongst other uh human rights

organizations and they wrote report

called xenophobic machines which really

highlighted what had happened here and

in effect they had developed on top of

bias data so there are already issues in

terms of reviewing these claims the

people who were using the software were

not fully trained and the developers

didn’t explain how it made its decisions

so complaints couldn’t be addressed what

is the compelling reason to use uh this

software and their compelling reason to

use it was that they would be

incentivized if they created more

clawback

demands it was not a virtuous circle

should we say and that’s probably one of

the worst cases the market is moving

really really rapidly at the moment and

chat GPT obviously made a really really

large Splash in the last few years as

that’s maturing what we’re seeing more

and more as AI tools becoming less of a

bespoke thing you are accessing and more

something that’s part and parcel of what

you already have in your pocket every

day and that is only going to increase

that effect that you can simply ask

you’ve been asked to do a risk

assessment to make it very practical for

most Risk Managers you’ve sent out a

request to do a risk identification

project nine times out of 10 we have to

expect that human nature means that

individuals are going to be using these

tools in their pockets to help them in

their workflow to help them answer that

question what was typically a

whiteboarding and a paper based process

and and Years Gone by is going to be

that but the challenges for Risk

Managers is how they make sure that

their internal tools and systems and the

access that’s there can be equivalently

accessible to staff in a secure

environment so they can access that kind

of knowledge and workflow that they have

in their pocket in a secure environment

but also that the models that they’re

using are informed by internal data so

they’re not basing you know Bas Baseline

information that they’re using from AI

systems to help them answer basic

questions and improve their workflows is

based on internal data from the

organization rather than just generic

what check GPT is giving them what is

soon to be giving them um all the likes

of perplexity Google and Microsoft’s

products I suppose in a danger there

there’s a lot of Silent integration

going on right and we’ll soon lose the

ability to identify where where AI is

doing things and uh what is a baseline

Etc yeah one of the interesting things

that we looked at was inde deep fake

technology and again there’s some really

widely now known cases where there has

been significant fraud

where we’re now starting to see where we

used to see sort of fishing attempts on

an email basis now we can actually start

to do that over a call to give you

example in the system I’m running right

now I will make a change and you’ll see

my camera will stop for just a second it

will come back on and as it does I am

now in real time on the call with you I

am somebody else you may recognize this

person but it is not difficult for me to

change over and over now a couple things

you’ll recognize pretty quickly my voice

is still the same I haven’t changed my

voice now that can be done it’s a little

bit more complicated more than we want

to get into today but to sync a

different voice and a different face

together is now possible and it’s it is

doable with with a very small delay but

my hair hasn’t changed my clothing

hasn’t changed so the system itself

really still needs to map to some facial

structure that’s similar to the person

you’re trying to mimic it’s helpful if

you can mimic their sort of motions the

way they smile the way they act the way

they talk in order to convince somebody

to do something now most people will

tell me on a deep fake immediately I

well I wouldn’t be fooled by that

because I I’m going to recognize this is

a deep fake but you have to put yourself

into a position of a very large

organization where you may not

personally know some senior Executives

so you don’t know their mannerisms don’t

know the kinds of contextual clues that

you would be able to pick up and if they

tell you or ask you to do something in a

call you’re in some cases likely to do

so can I ask each panelists where they

think AI is headed definity AGI

everybody is talking about with is

artificial general intelligence I I I

don’t I hate that name I could call it

augmented general intelligence because I

think uh human will be still there I

more into leak lugger philosophy that it

will be symbiosis between man achine

with every Pro

progress very very risk and and uh

Danger from what I’m hearing and

understanding from smarter people is

that these huge computational leaps uh

might be resting a little bit so the

computer power that was necessary to

produce chat gbt or the big llms that

might be resting a little bit meaning

that we won’t see these giant leaps in

the future but maybe in increased

development and adaptability and

Adoption of these Technologies right

also as Peter is indicating here that we

have the maybe the existing

toolbox but they will become more and

more uh Advanced and more and more

integrated as um into our daily life and

into becoming tools that we use uh

working and privately and whatever we

might also see a reduction in the

skepticism towards new technology it has

always been there and maybe that will

maybe Flatline a little bit and see okay

we can actually use the tool in our

everyday life in the near term over the

next sort of two years or so we’re going

to see a lot of organizations trying and

figure out how to use this technology

which is going to drive the costs up we

don’t have what we need for industry as

to how much the demand is going to be we

going to start to see organizations

really have to figure out okay what’s am

I going to invest in more digital sort

of power digital uh capital or am I

going to invest into human capital and

every day will be a new Balancing Act as

we start to sort of lean into Quantum

Computing which will again change things

over the next sort of fiveyear horizon

over a 10year horizon we’ll start to

realize we get sort of energy becomes a

bit more abundant we’re going to start

to see things like fusion and other so

we’ll solve some of the elements that

we’re starting to see capacity

constraints that we’ll we’ll experience

now but I think a lot of the biggest

challenges now are just hitting those

capacity constraints as sort of was

mentioned already with this plateau and

how are we going to manage those um you

know organization by organization and

how are we going to fund the development

the continued development whether it’s

nuclear energy or others how are we

going to fund the development of these

technologies that we all want to use the

competition’s very quickly switching to

organizations getting access whether or

not that’s talent whether or not that’s

with key Partnerships and strategic

Partnerships to lock down their

exclusive access against competitors to

that kind of model or to build that

internally and the resources that that’s

going to take what we’re already seeing

is a little bit of a need reaction

around implementation towards uh

efficiency and headcount reduction it’s

it’s the broader economic environment as

well and my sense is that one of the

things we’re going to see is an

increasing realization that these are

augmenting what your staff are doing and

the knowledge that you have across your

sites and across your staff Network

rather than replacing that I would like

to say thank you very much to Pauline

Hart mads Peter and Doug uh and thank

you David ruin for uh obviously setting

up the Network the isrm in the first

place and uh helping us uh get this

conversation going out there into the

wider world uh thank you one and all

those of you that joined us and I hope

you find the recording very useful

thanks and from Lond