WEBVTT
00:00:00.479 --> 00:00:02.799
I'm Robin Amler of IBS Intelligence.
00:00:02.879 --> 00:00:05.679
You're listening to the IBSI Views podcast.
00:00:05.839 --> 00:00:18.399
With me is Rob Mason, Director of Regulatory Intelligence at Global Relay, which recently put out a report saying that 66% of financial services firms now ban communications channels.
00:00:18.960 --> 00:00:21.440
Why are channels being banned, Rob?
00:00:21.679 --> 00:00:23.280
What's going on here?
00:00:24.079 --> 00:00:25.440
So thanks very much, Robin.
00:00:25.600 --> 00:00:26.559
Lovely to be here.
00:00:26.719 --> 00:00:39.759
Just to start by saying that financially regulated firms have a reg requirement to deliver and keep relevant records for respective communications to monitor them against a number of risks, and primarily that's around market abuse.
00:00:39.840 --> 00:00:45.600
So things like insider dealing, where they relate to orders and trades in a regulated asset class.
00:00:45.759 --> 00:01:02.719
And what we saw is the US take a very strong regulatory view on those record keeping requirements, especially around some particular channels, WhatsApp being one that was named quite frequently with massive fines across the industry because some channels were not captured.
00:01:03.039 --> 00:01:08.079
So those communication channels were required to be prohibited.
00:01:08.640 --> 00:01:16.000
And in some degree, that had always happened, and firms maintain a list of approved channels which are recorded and captured for compliance.
00:01:16.159 --> 00:01:20.640
However, prohibition clearly doesn't work as a risk management strategy.
00:01:20.799 --> 00:01:23.760
So the expectation is that firms need to do more.
00:01:24.000 --> 00:01:35.519
So as a vendor, for example, that we work for at Global Relay, we've developed solutions that can enable some of those more esoteric communications channel communication channels to be used.
00:01:35.680 --> 00:01:40.560
And we feel we enable firms to operate more widely and do so compliantly.
00:01:40.879 --> 00:01:52.079
Banning of those channels is still in play at some firms, but there's a much clearer understanding of the risks, unsurprisingly, when you consider$3 billion worth of fines across the industry.
00:01:52.159 --> 00:01:54.959
So that has the uh habit of focusing the mind a little bit.
00:01:55.040 --> 00:01:58.799
And we've seen a lot of enhanced focus and attention in this space as a result.
00:01:59.280 --> 00:02:00.560
You mentioned WhatsApp.
00:02:00.719 --> 00:02:03.599
How are you going to know if I'm on WhatsApp?
00:02:04.079 --> 00:02:05.519
So it's a it's a great question.
00:02:05.680 --> 00:02:11.360
I think it's more really to make sure that everybody understands the consequences of their actions.
00:02:11.759 --> 00:02:34.639
So, generally speaking, and in the early stages of when this was beginning to unravel and the SEC, the Securities Exchange Commission in the US took a fairly dim view of this, some firms decided to try and recoup some of the fines that were being levied against them by making people personally liable and they were essentially sort of docking bonuses or wages as a consequence.
00:02:34.960 --> 00:02:58.319
So, what that is, is if your client or the person on the end of the phone who wants to deal with you chooses to use WhatsApp, it's the responsibility of the employee at the organization to say, no, I can't do business on this particular channel and give you an alternative, or make efforts to be able to record or capture that so that it can be fed into the respective systems for compliance.
00:02:58.560 --> 00:03:17.120
Now, that's not to say that you can jump out of the office and have a WhatsApp conversation on the side without anybody knowing, and I think that's a prevailing risk that everybody's aware of and has been in the course of history, particularly around things like insider dealing, where obviously dissemination or or spreading of information is key.
00:03:17.280 --> 00:03:18.080
You can't manage that.
00:03:18.159 --> 00:03:31.039
But from a firm perspective, you need to be able to do everything in your power to be able to train and make sure everyone else is aware what is permitted and what sits within the rules and regulations of the business that they're undertaking.
00:03:31.520 --> 00:03:34.400
This is about governance, it's about compliance.
00:03:35.199 --> 00:03:41.759
But there is an issue of, I understand, regional differences in the way that this is actually handled.
00:03:42.080 --> 00:03:48.000
It's different in North America to the way it's handled in Europe, to the way it's handled in Asia.
00:03:48.240 --> 00:03:51.120
Do we need glob your global relay?
00:03:51.280 --> 00:03:53.360
Do we need global solutions?
00:03:53.919 --> 00:03:55.360
Yes, it's an interesting one.
00:03:55.439 --> 00:04:05.919
I mean, uh the bigger institutions have a generally adopt a global minimum standard, and what they take is the most sort of strict regulation and try and adopt that.
00:04:06.000 --> 00:04:12.879
And if they don't do that and they have regional regulations, then it forms some cracks and things tend to fall down those cracks.
00:04:13.120 --> 00:04:24.639
I spent seven years working at a big Swiss investment bank as the head of their monitoring and surveillance piece, and we decided to go down the path of how of having a global minimum standard across the board.
00:04:24.800 --> 00:04:33.759
Now that might have exceeded some of the local regulations in some of the jurisdictions in which we operated, but we felt much more comfortable in adopting that and a consistent line.
00:04:33.839 --> 00:04:36.480
And I think consistency is the key.
00:04:36.639 --> 00:04:48.959
Now, with regards to North America, yeah, I mean, interestingly, they themselves probably were the subject of the majority of the regul of the regulators, the SEC's focus.
00:04:49.120 --> 00:04:59.360
So consequently, those punitive fines meant that they've sort of taken probably the most radical action to make sure that they don't suffer the same fate in future.
00:04:59.839 --> 00:05:02.480
Once bitten twice shy, perhaps.
00:05:02.800 --> 00:05:03.600
Yeah, absolutely.
00:05:03.680 --> 00:05:20.639
I mean, as I said, uh across a period of about three years, I think it kind of the total's a bit variable, but across industry and across firm, not a single firm, but across all of the firms which the SEC captured, it was something in the vicinity of three billion with a B billion dollar fines.
00:05:20.800 --> 00:05:31.360
So you can imagine that there was a fairly sort of significant ripple in the pool with regards to those folks that were paying those fines and making sure everybody knew that that wasn't going to happen again.
00:05:31.920 --> 00:05:42.879
Now there's a balance, or if not balance, a pendulum between monitoring and surveillance on one side and an outright ban on the other.
00:05:44.000 --> 00:05:45.279
Where should we be?
00:05:45.439 --> 00:05:46.639
What what works?
00:05:46.800 --> 00:05:49.120
What what's going to be most effective?
00:05:49.680 --> 00:05:57.839
Yeah, I as I said at the top, you know, there's always been this list of communication channels which are approved and some which are not approved.
00:05:57.920 --> 00:06:00.959
So the non-approved ones are essentially prohibited from use.
00:06:01.120 --> 00:06:03.279
But clearly that doesn't work.
00:06:03.519 --> 00:06:07.839
That's merely a prohibition, but it's not a risk management solution.
00:06:08.079 --> 00:06:11.519
So what was required is that yes, you do need to have that list.
00:06:11.600 --> 00:06:28.720
And despite the fact that more channels are being enabled because of some of the vendor work around this and making sure that business can be enabled across a wider variety of channels, there is nevertheless an implicit requirement for the regulated community to do more than that.
00:06:28.879 --> 00:06:50.079
So what we've seen is some banks and brokers, particularly the bigger ones where the larger risks prevail, they might themselves send like a phishing message, like you would in your email, but maybe a text message on WhatsApp, just to test to see if any individual picks that up and then doesn't report it or doesn't appropriately deal with that and suggest that there may be some business there.
00:06:50.240 --> 00:07:00.800
So we've seen a different some different methodologies, but it's more about banning is appropriate, but it shouldn't be in isolation because it doesn't manage the risk, it just tells you what the rules are.
00:07:01.199 --> 00:07:11.839
It's a dynamic situation that is continually evolving, not least, and we haven't mentioned this yet, and I'm going to bring it into the conversation, artificial intelligence.
00:07:12.480 --> 00:07:20.079
Because there's a whole other set of issues there about what AIs get up to and what agents get up to on their own.
00:07:20.399 --> 00:07:25.360
So I I mean, good time to mention this, and you know, very much worth talking about this.
00:07:25.519 --> 00:07:27.519
I mean, some positives first, perhaps.
00:07:27.600 --> 00:07:42.079
So the way that banks and brokers and the regulated community manage this risk in terms of identifying some issues that might be in a communication, um, historically has been performed using lexicons, and these are effectively word lists.
00:07:42.240 --> 00:07:51.759
So when that word that was deemed suspicious crops up in an email exactly, it would flag for investigation of an analyst.
00:07:51.839 --> 00:08:00.000
Now, that's a pretty blunt instrument, and this has resulted in a huge amount of what they describe as false positives and rarely any real risk detected.
00:08:00.240 --> 00:08:02.319
And AI has completely changed this.
00:08:02.399 --> 00:08:14.480
So the way that the AI has leveraged is it will review a communication in that in totality and it will consider whether there's been reasonable grounds for suspicion and flag that accordingly, and it does so incredibly effectively.
00:08:14.720 --> 00:08:23.439
This has resulted in massive reduction in those in those false positives, but also a huge uptick in the number of truly suspicious communications.
00:08:23.759 --> 00:08:35.919
Now, some of the issues that the the big institutions in particular are facing is that people are using um large language models like Copilot and ChatGPT and Claude in their everyday work.
00:08:36.080 --> 00:08:42.559
And what are the regulations or rules around that and how can they maintain some guidelines is very difficult.
00:08:42.799 --> 00:08:46.799
As ever, there'll be lots of governance around that, so there's lots of good advice.
00:08:47.039 --> 00:09:04.000
But what we've suggested that we can we can help them with is we provide um a uh a communication ability to be able to capture anything that's fed into one of those types of of um large language models.
00:09:04.080 --> 00:09:10.559
So we call it a connector, and what that does is that collects all the information that may be pushed into a chat GPT.
00:09:10.639 --> 00:09:17.600
Now you may or not want to monitor that, but what you can do then is if you have another suspicion, you can look to see what potentially an employee has put in there.
00:09:17.759 --> 00:09:23.759
And it might be something very, very um innocuous, like could you redraft this email in a certain style or with bullets?
00:09:23.840 --> 00:09:32.639
But it might be something that's got sensitive data or client-identifying data or something in it, which a bank and broker is not going to want to share with a wider audience.
00:09:32.720 --> 00:09:40.639
And by putting it into one of those large language models, it's very difficult to know how restricted that information is or whether it's open source for anybody to have a look at.
00:09:41.039 --> 00:09:47.120
If you put it into an LLM, I would argue that you're basically publishing.
00:09:47.519 --> 00:09:50.720
Well, this is the problem, and I think the banks share this view.
00:09:50.879 --> 00:09:58.480
And whilst you know you can have some lots of training and lots of guidance, similarly, you can't account for what an employee might do.
00:09:58.559 --> 00:10:10.000
And generally, it might not even be malicious, but it might be just ignorance that someone pops something in there because they're not quite sure of a question, and with it they share some sensitive information that could therefore be accessed by another.
00:10:10.159 --> 00:10:17.600
I guess probably it's more of a sort of a CISO, like a cyber security type problem, rather than anything from a market abuse perspective.
00:10:17.679 --> 00:10:23.679
But nevertheless, the banks and brokers take this very seriously because all of that data is super sensitive or a significant amount of it.
00:10:24.159 --> 00:10:27.360
So just to round up, what's global relay itself doing?
00:10:27.600 --> 00:10:29.039
What do you have on offer?
00:10:29.200 --> 00:10:31.600
How do you solve this problem for people?
00:10:32.399 --> 00:10:42.559
Yeah, so I think for our initial use case, we're a record keeping and um a surveillance, monitoring, and surveillance of communications solutions vendor.
00:10:42.720 --> 00:10:48.399
We've got a huge number of clients signing in the region of excess of 10,000 clients, been operating in the business for 25 years.
00:10:48.480 --> 00:10:49.919
So we know what we're doing here.
00:10:50.080 --> 00:10:57.759
And the way that we're identifying some of the risks that might be prevailing is a different way to most of our competitors, and it is leveraging large language models.
00:10:57.919 --> 00:11:05.919
Now there's some very, very clever stuff around things like risk identification, but also translation and transcription.
00:11:06.000 --> 00:11:22.080
So pretty much translation of any language, and therefore that brings also into uh transcription of any voice communication, and this is a huge improvement from what was previously on offer and allows voice communications to also be incorporated into a monitoring program for capture.
00:11:22.480 --> 00:11:40.879
Um what we're seeing is an enormous appetite for this to be undertaken because I think, as I said, that the previous system was relatively ineffective, and we're seeing a lot of the institutions across the board, all regulated asset classes, all regulated types, shapes and sizes in financial services, looking to come and have a conversation about this latest technology.
00:11:40.960 --> 00:11:52.720
And interestingly, too, even the regulators want to come and have a chat with us to understand what it's on offer so that they can look at their regulated community to ensure that they're keeping up with managing the risks in the best way that they can.
00:00:00.479 --> 00:00:02.799
I'm Robin Amler of IBS Intelligence.
00:00:02.879 --> 00:00:05.679
You're listening to the IBSI Views podcast.
00:00:05.839 --> 00:00:18.399
With me is Rob Mason, Director of Regulatory Intelligence at Global Relay, which recently put out a report saying that 66% of financial services firms now ban communications channels.
00:00:18.960 --> 00:00:21.440
Why are channels being banned, Rob?
00:00:21.679 --> 00:00:23.280
What's going on here?
00:00:24.079 --> 00:00:25.440
So thanks very much, Robin.
00:00:25.600 --> 00:00:26.559
Lovely to be here.
00:00:26.719 --> 00:00:39.759
Just to start by saying that financially regulated firms have a reg requirement to deliver and keep relevant records for respective communications to monitor them against a number of risks, and primarily that's around market abuse.
00:00:39.840 --> 00:00:45.600
So things like insider dealing, where they relate to orders and trades in a regulated asset class.
00:00:45.759 --> 00:01:02.719
And what we saw is the US take a very strong regulatory view on those record keeping requirements, especially around some particular channels, WhatsApp being one that was named quite frequently with massive fines across the industry because some channels were not captured.
00:01:03.039 --> 00:01:08.079
So those communication channels were required to be prohibited.
00:01:08.640 --> 00:01:16.000
And in some degree, that had always happened, and firms maintain a list of approved channels which are recorded and captured for compliance.
00:01:16.159 --> 00:01:20.640
However, prohibition clearly doesn't work as a risk management strategy.
00:01:20.799 --> 00:01:23.760
So the expectation is that firms need to do more.
00:01:24.000 --> 00:01:35.519
So as a vendor, for example, that we work for at Global Relay, we've developed solutions that can enable some of those more esoteric communications channel communication channels to be used.
00:01:35.680 --> 00:01:40.560
And we feel we enable firms to operate more widely and do so compliantly.
00:01:40.879 --> 00:01:52.079
Banning of those channels is still in play at some firms, but there's a much clearer understanding of the risks, unsurprisingly, when you consider$3 billion worth of fines across the industry.
00:01:52.159 --> 00:01:54.959
So that has the uh habit of focusing the mind a little bit.
00:01:55.040 --> 00:01:58.799
And we've seen a lot of enhanced focus and attention in this space as a result.
00:01:59.280 --> 00:02:00.560
You mentioned WhatsApp.
00:02:00.719 --> 00:02:03.599
How are you going to know if I'm on WhatsApp?
00:02:04.079 --> 00:02:05.519
So it's a it's a great question.
00:02:05.680 --> 00:02:11.360
I think it's more really to make sure that everybody understands the consequences of their actions.
00:02:11.759 --> 00:02:34.639
So, generally speaking, and in the early stages of when this was beginning to unravel and the SEC, the Securities Exchange Commission in the US took a fairly dim view of this, some firms decided to try and recoup some of the fines that were being levied against them by making people personally liable and they were essentially sort of docking bonuses or wages as a consequence.
00:02:34.960 --> 00:02:58.319
So, what that is, is if your client or the person on the end of the phone who wants to deal with you chooses to use WhatsApp, it's the responsibility of the employee at the organization to say, no, I can't do business on this particular channel and give you an alternative, or make efforts to be able to record or capture that so that it can be fed into the respective systems for compliance.
00:02:58.560 --> 00:03:17.120
Now, that's not to say that you can jump out of the office and have a WhatsApp conversation on the side without anybody knowing, and I think that's a prevailing risk that everybody's aware of and has been in the course of history, particularly around things like insider dealing, where obviously dissemination or or spreading of information is key.
00:03:17.280 --> 00:03:18.080
You can't manage that.
00:03:18.159 --> 00:03:31.039
But from a firm perspective, you need to be able to do everything in your power to be able to train and make sure everyone else is aware what is permitted and what sits within the rules and regulations of the business that they're undertaking.
00:03:31.520 --> 00:03:34.400
This is about governance, it's about compliance.
00:03:35.199 --> 00:03:41.759
But there is an issue of, I understand, regional differences in the way that this is actually handled.
00:03:42.080 --> 00:03:48.000
It's different in North America to the way it's handled in Europe, to the way it's handled in Asia.
00:03:48.240 --> 00:03:51.120
Do we need glob your global relay?
00:03:51.280 --> 00:03:53.360
Do we need global solutions?
00:03:53.919 --> 00:03:55.360
Yes, it's an interesting one.
00:03:55.439 --> 00:04:05.919
I mean, uh the bigger institutions have a generally adopt a global minimum standard, and what they take is the most sort of strict regulation and try and adopt that.
00:04:06.000 --> 00:04:12.879
And if they don't do that and they have regional regulations, then it forms some cracks and things tend to fall down those cracks.
00:04:13.120 --> 00:04:24.639
I spent seven years working at a big Swiss investment bank as the head of their monitoring and surveillance piece, and we decided to go down the path of how of having a global minimum standard across the board.
00:04:24.800 --> 00:04:33.759
Now that might have exceeded some of the local regulations in some of the jurisdictions in which we operated, but we felt much more comfortable in adopting that and a consistent line.
00:04:33.839 --> 00:04:36.480
And I think consistency is the key.
00:04:36.639 --> 00:04:48.959
Now, with regards to North America, yeah, I mean, interestingly, they themselves probably were the subject of the majority of the regul of the regulators, the SEC's focus.
00:04:49.120 --> 00:04:59.360
So consequently, those punitive fines meant that they've sort of taken probably the most radical action to make sure that they don't suffer the same fate in future.
00:04:59.839 --> 00:05:02.480
Once bitten twice shy, perhaps.
00:05:02.800 --> 00:05:03.600
Yeah, absolutely.
00:05:03.680 --> 00:05:20.639
I mean, as I said, uh across a period of about three years, I think it kind of the total's a bit variable, but across industry and across firm, not a single firm, but across all of the firms which the SEC captured, it was something in the vicinity of three billion with a B billion dollar fines.
00:05:20.800 --> 00:05:31.360
So you can imagine that there was a fairly sort of significant ripple in the pool with regards to those folks that were paying those fines and making sure everybody knew that that wasn't going to happen again.
00:05:31.920 --> 00:05:42.879
Now there's a balance, or if not balance, a pendulum between monitoring and surveillance on one side and an outright ban on the other.
00:05:44.000 --> 00:05:45.279
Where should we be?
00:05:45.439 --> 00:05:46.639
What what works?
00:05:46.800 --> 00:05:49.120
What what's going to be most effective?
00:05:49.680 --> 00:05:57.839
Yeah, I as I said at the top, you know, there's always been this list of communication channels which are approved and some which are not approved.
00:05:57.920 --> 00:06:00.959
So the non-approved ones are essentially prohibited from use.
00:06:01.120 --> 00:06:03.279
But clearly that doesn't work.
00:06:03.519 --> 00:06:07.839
That's merely a prohibition, but it's not a risk management solution.
00:06:08.079 --> 00:06:11.519
So what was required is that yes, you do need to have that list.
00:06:11.600 --> 00:06:28.720
And despite the fact that more channels are being enabled because of some of the vendor work around this and making sure that business can be enabled across a wider variety of channels, there is nevertheless an implicit requirement for the regulated community to do more than that.
00:06:28.879 --> 00:06:50.079
So what we've seen is some banks and brokers, particularly the bigger ones where the larger risks prevail, they might themselves send like a phishing message, like you would in your email, but maybe a text message on WhatsApp, just to test to see if any individual picks that up and then doesn't report it or doesn't appropriately deal with that and suggest that there may be some business there.
00:06:50.240 --> 00:07:00.800
So we've seen a different some different methodologies, but it's more about banning is appropriate, but it shouldn't be in isolation because it doesn't manage the risk, it just tells you what the rules are.
00:07:01.199 --> 00:07:11.839
It's a dynamic situation that is continually evolving, not least, and we haven't mentioned this yet, and I'm going to bring it into the conversation, artificial intelligence.
00:07:12.480 --> 00:07:20.079
Because there's a whole other set of issues there about what AIs get up to and what agents get up to on their own.
00:07:20.399 --> 00:07:25.360
So I I mean, good time to mention this, and you know, very much worth talking about this.
00:07:25.519 --> 00:07:27.519
I mean, some positives first, perhaps.
00:07:27.600 --> 00:07:42.079
So the way that banks and brokers and the regulated community manage this risk in terms of identifying some issues that might be in a communication, um, historically has been performed using lexicons, and these are effectively word lists.
00:07:42.240 --> 00:07:51.759
So when that word that was deemed suspicious crops up in an email exactly, it would flag for investigation of an analyst.
00:07:51.839 --> 00:08:00.000
Now, that's a pretty blunt instrument, and this has resulted in a huge amount of what they describe as false positives and rarely any real risk detected.
00:08:00.240 --> 00:08:02.319
And AI has completely changed this.
00:08:02.399 --> 00:08:14.480
So the way that the AI has leveraged is it will review a communication in that in totality and it will consider whether there's been reasonable grounds for suspicion and flag that accordingly, and it does so incredibly effectively.
00:08:14.720 --> 00:08:23.439
This has resulted in massive reduction in those in those false positives, but also a huge uptick in the number of truly suspicious communications.
00:08:23.759 --> 00:08:35.919
Now, some of the issues that the the big institutions in particular are facing is that people are using um large language models like Copilot and ChatGPT and Claude in their everyday work.
00:08:36.080 --> 00:08:42.559
And what are the regulations or rules around that and how can they maintain some guidelines is very difficult.
00:08:42.799 --> 00:08:46.799
As ever, there'll be lots of governance around that, so there's lots of good advice.
00:08:47.039 --> 00:09:04.000
But what we've suggested that we can we can help them with is we provide um a uh a communication ability to be able to capture anything that's fed into one of those types of of um large language models.
00:09:04.080 --> 00:09:10.559
So we call it a connector, and what that does is that collects all the information that may be pushed into a chat GPT.
00:09:10.639 --> 00:09:17.600
Now you may or not want to monitor that, but what you can do then is if you have another suspicion, you can look to see what potentially an employee has put in there.
00:09:17.759 --> 00:09:23.759
And it might be something very, very um innocuous, like could you redraft this email in a certain style or with bullets?
00:09:23.840 --> 00:09:32.639
But it might be something that's got sensitive data or client-identifying data or something in it, which a bank and broker is not going to want to share with a wider audience.
00:09:32.720 --> 00:09:40.639
And by putting it into one of those large language models, it's very difficult to know how restricted that information is or whether it's open source for anybody to have a look at.
00:09:41.039 --> 00:09:47.120
If you put it into an LLM, I would argue that you're basically publishing.
00:09:47.519 --> 00:09:50.720
Well, this is the problem, and I think the banks share this view.
00:09:50.879 --> 00:09:58.480
And whilst you know you can have some lots of training and lots of guidance, similarly, you can't account for what an employee might do.
00:09:58.559 --> 00:10:10.000
And generally, it might not even be malicious, but it might be just ignorance that someone pops something in there because they're not quite sure of a question, and with it they share some sensitive information that could therefore be accessed by another.
00:10:10.159 --> 00:10:17.600
I guess probably it's more of a sort of a CISO, like a cyber security type problem, rather than anything from a market abuse perspective.
00:10:17.679 --> 00:10:23.679
But nevertheless, the banks and brokers take this very seriously because all of that data is super sensitive or a significant amount of it.
00:10:24.159 --> 00:10:27.360
So just to round up, what's global relay itself doing?
00:10:27.600 --> 00:10:29.039
What do you have on offer?
00:10:29.200 --> 00:10:31.600
How do you solve this problem for people?
00:10:32.399 --> 00:10:42.559
Yeah, so I think for our initial use case, we're a record keeping and um a surveillance, monitoring, and surveillance of communications solutions vendor.
00:10:42.720 --> 00:10:48.399
We've got a huge number of clients signing in the region of excess of 10,000 clients, been operating in the business for 25 years.
00:10:48.480 --> 00:10:49.919
So we know what we're doing here.
00:10:50.080 --> 00:10:57.759
And the way that we're identifying some of the risks that might be prevailing is a different way to most of our competitors, and it is leveraging large language models.
00:10:57.919 --> 00:11:05.919
Now there's some very, very clever stuff around things like risk identification, but also translation and transcription.
00:11:06.000 --> 00:11:22.080
So pretty much translation of any language, and therefore that brings also into uh transcription of any voice communication, and this is a huge improvement from what was previously on offer and allows voice communications to also be incorporated into a monitoring program for capture.
00:11:22.480 --> 00:11:40.879
Um what we're seeing is an enormous appetite for this to be undertaken because I think, as I said, that the previous system was relatively ineffective, and we're seeing a lot of the institutions across the board, all regulated asset classes, all regulated types, shapes and sizes in financial services, looking to come and have a conversation about this latest technology.
00:11:40.960 --> 00:11:52.720
And interestingly, too, even the regulators want to come and have a chat with us to understand what it's on offer so that they can look at their regulated community to ensure that they're keeping up with managing the risks in the best way that they can.
