Automatically monitor regulatory updates to map to your internal policies, procesures and controls. Learn More
Automatically monitor regulatory updates to map to your internal policies, procesures and controls. Learn More

Speakers:
Asif Alam, CEO of Compliance.ai and Commodities Expert, Matthew Hunter

Webinar Details: The capacity of both the regulated and the regulator’s abilities to detect potential violations. Will address ideas about data sources and their limitations, the likely uses of AI to help regulators and enforcement agencies, and private firms’ compliance programs.

  • Existing regulations and data limitations
  • Spot markets and swaps structural differences including cryptocurrency or digital assets
  • Possible data collection regulations
  • The potential of AI
  • Regulated entity benchmark manipulation monitoring enhancement

Ronjini Joshua <Moderator>

Hello everybody, thank you for joining us. We’ll get started in a few minutes, we’re just waiting for people to file in, and we’ll get started.

Introduction

Moderator

Okay. Hello and welcome to Compliance AI’s webinar: “Regulators’ Capacity to detect market interference and Manipulation.” This will be a discussion between our Compliance AI CEO, Asif Alam, and our Compliance AI commodities expert, Matthew Hunter. I will be handing it over to Asif in a moment. If you have any Q & A’s during the discussion, then please feel free to dump them in the chat box and at the end we’ll have a short q&a session where we can answer any additional questions that come up from this conversation. So, Asif, please take it away.

Asif Alam <Asif>

Thank you very much. Ronjini. Hi, guys. I’m Asif Alam, Chief Executive Officer at Compliance AI. Good morning. Good afternoon. Good evening, wherever you are. It’s a pleasure for me to have Matthew Hunter with us in here who is truly an expert in this area. Matthew, it would be great if you take a few minutes to introduce yourself.

Matthew Hunter <Matthew>

Sure, sure thing. First a minor correction: I am not an AI expert. I am an opinionated person on AI with some experience with it, but I’m not an expert. I am an expert in physical and derivative commodities trading. I was a professional trader (Wall Street-trained) in quite a number of commodities in New York City. I have been involved in the regulatory side with two agencies, one of which I was the subject matter expert that provided guidance that ultimately created the division of analytics and surveillance at the FERC, the Federal Energy Regulatory Commission, and I was the deputy director of the Division of Enforcement surveillance branch at the CFTC for about 10 years. And I’ve been retired for roughly two years.

Asif 

Excellent. And a little bit about myself. So, as I said, I’m the CEO of Compliance AI, but prior to being the CEO of Compliance, I worked at the company Thomson Reuters for a very long time (for about 18 plus years) then I did a few startups. My expertise primarily has been (more or less) on data and AI. Back in the 2000s, I started working with AI. Prior to this company that was very much focused on AI and here at Compliance AI, we do a lot of AI. So throughout my career, even despite different products and product sets, mostly focusing on banking and financial institutions in FinTech, legal tech and REG tech, I’ve been doing different kinds of AI for that matter. So again, welcome, everybody, and really looking forward to having a great discussion here with Matthew. These are some of the topics that we will be discussing.

<To Moderator> (Ronjini, we can leave it leave the screen in there,) but let’s get right into it.

Matthew, as we live in this data-driven world, can regulators and firms see violations?

Matthew

That’s a great question. So, as in everything, the world is nuanced. Remember, most of my focus is not on SEC jurisdictional products, but on CFTC jurisdictional products to the commodities world (and that includes swaps and other derivatives.) The question is, “Do you have the available information?” The history of the CFTC was that the staff was examining futures trading. If you only look at futures trading, then you could say that they have a “complete picture” of the world because the data is very robust and has been for a long time related to futures. But if you say that the market is more complex, and I say the market is more complex, because having traded, you have the physical world, you have had the futures world and other derivatives. So effectively, you have three stools. If you only look at one leg of the stool, then you have a very narrow perspective of the world. And in my mind view, all regulators have a fairly constrained view, they don’t have an open view of the world.

Asif 

So, let me go back to a question around data, Matthew, (for everybody who’s listening to this), what is your definition of a “full set of data” if you will?

Matthew

So, for me, the full set of data is all data related to a product. So, in commodities, that means all physical trades, all futures-related trades, and that’s Futures and Options on futures, and all swamps. If you only have a portion of it, you cannot say you understand what is happening in not only the behavior of an actor or actors, but you cannot say you understand the market, because you’re only looking at pieces. It’s rather like the frog that’s at the bottom of a well looks up and says I have the best piece of sky, I have the best sky because it can see through here up from the bottom of the well, the blue sky, but it doesn’t see all of the sky and it needs to see to experience the world, the full sky. So, a full data set is something that is rather complete. That’s my view.

Asif

So, Matthew are you implying that a firm’s own data is superior somehow to what is provided to regulators in, sort of, the financial and physical spaces?

Matthew

Yes, I am saying that a firm can see a perfect vision of its own participation in the market. It can know how big it is in relation to that market. It can know what its contributions are and it can assess more accurately how it impacts price formation, than a regulator. A regulator and CFTC (at its best), has two legs of the stool. It has the derivative side, and it has the future side, but it doesn’t have the physical side. And if you think about DOJ and other agencies, there isn’t a central collection of trading. There isn’t any central depository or repository for the physical or the cash side of trading. There are only places where you have swaps being reported, or you have futures and futures trading being reported. But you don’t see the suite. So, in commodities or in indices that are shared by more than one trading venue, each trading venue might be able to see what’s happening on its platform, but it cannot see the complete picture and neither can a regulator.

Asif

So, does that mean that there are differences between (sort of) the spot market, especially the thing that is so hot right now; cryptocurrencies and digital assets (and other markets) from a data perspective? Really, as we stay on data.

Matthew

Yes, absolutely, so there isn’t any collection. I think this is where Chairman Gensler has a problem. And the problem is there isn’t any real regulation of cash market trading platforms. The NFA recently tried or is trying to place on its members an obligation to conform to its rules for trading futures (for those members that are trading, and I don’t want to misspeak),  I think it’s only related to retail, but to the retail participation in cryptocurrencies. There are no regulations on natural gas trading, there aren’t any regulations, particularly about and centralized collection of data for Treasury or foreign exchange, gold, platinum, and copper. That type of cash market oversight doesn’t exist in a way that is akin to the way derivatives are supervised. So, the trades aren’t reported anywhere, there isn’t access by call it “interested enforcement parties” to be able to do surveillance on that data. As an example, the CFTC does not have cash market data for any number of commodities, or access to it. So like the USDA data for agricultural commodities, there’s actually a prohibition right in the law for the USDA to provide that data for the purpose of surveillance. If there’s an investigation and a subpoena, that information can be provided, but not for looking for the potential abuse in a market. And so, I think it’s quite true for the majority of government regulators is that they don’t see the full picture of physical futures and other derivatives, they see pieces and from the pieces might have a concern and get more data. The CFTC certainly can get more data using a special call from their authority before subpoenas, and other agencies can use their authorities to get more complete data. But there isn’t any central collection to say I will reach it from like crypto a foreign exchange of gold by going to the repository depository itself.

Asif  

So, what I’m listening to from you, and just let me qualify that. Are there distinctions between retail trading platforms and wholesale trading platforms in the spot market?

Matthew

On a very broad stroke, the answer has to be yes. You know, retail has its own rules, crypto has pointed out some of the flaws and those rules, where transactions that would not be allowed in a derivative platform and the treatment of customer funds would not be allowed under the derivative rules. I think that points to there are lots of platforms is no regulator that says I am responsible for cash market trade, and you must apply to me to list the cash market. It’s not to say that there aren’t state rules, or other rules about the commodity. Obviously, Treasury has an interest in who can trade in its auction. And so the secondary market has rules and the participants themselves put on credit controls about trading with each other but a central meeting place as an exchange, you don’t have to register with anyone (to do cash trading transactions). A few years ago, there wasn’t one foreign exchange. So there are lots of retail foreign exchange markets. Just go on the internet and participate, though CFTC has punished, particularly in the retail metals area, and some in the foreign exchange area have punished misprints.

Asif

And I think I know the answer to what you will say. But I will still ask it for the audience: What data is provided to the regulators, and are all regulators equally situated in that sense?

Matthew

So, there’s data that’s obviously provided to the CFTC. And it’s of very high quality, there are two levels of that data. One is transaction-level data, which is what every market participant sees in their historical data. You see the buys and sells, volume and estimated volume, and the price with a timestamp. But there’s also something called message data, which is every order (and the structure to some degree) the structure of those orders and where those are in the queue (Second by second or millisecond by millisecond or nanosecond by nanosecond), And the change in orders themselves, plus the transactions that occur so that the regulator gets, CFTC gets, and that the SEC through FINRA has access to similar data, though I think the SEC problem is more complex, because there are any number of (I’ll use the word “exchanges”) that are listing the same products. And then there are those products that are the same, but trading on foreign forces. So, when you think about the ultimate enforcement actions by institutions like DOJ, they don’t have that data. It’s not collected, and so they would come to CFTC or SEC, to get data related to an interest that they’re pursuing. But there isn’t an independent place to get called the “third light.” Though the SEC is a little different than “What is the cash market?” doesn’t fit the same profiles or commodities.

Asif

I think one thing that I’m seeing Matthew is that, again, (I think you are also alluding to) is that if you look at the market, the complexity and the velocity when it comes to regulations and compliance overall, it’s just increasing, and it’s increasing proportionately. The numbers are just getting pretty big in that sense. What legislation (in that sense), would you propose to be enacted that enables regulators to detect violation?

Matthew

In my opinion, (because there is one level of cash transactions that are not recorded anywhere), all transactions should be recorded and placed in a central repository for, call it “government entities” to be able to reach the full data suite. So that special call I alluded to before is unnecessary because CFTC or another enforcement agency could see the complete picture of a market: every transaction, every bid and offer. So that price formation is understood completely. Market participation, size of the market is completely understood. And so, I think the primary rule is a reporting requirement, that if you trade (and obviously it would be some distinction about de minimis volumes and what is retail reporting to retail transaction, separately.) So that the entire entirety of a market (whatever that is), is accessible, quickly to a regulator. There were some attempts to add in 2008, the idea of bringing the swap market into the SEC or CFTC jurisdiction so that data would be reported somewhere and that you would have access to it. The idea is to bring in the last step, at least for the commodity world, it’s to bring in the physical.

Asif

Yes, Matthew, but records of transactions, orders and price formations for orders are required to be kept. How is this data accessed today? What happens there?

Matthew

Okay, so on its best day when an analysis of an event takes place, (and I call it that from surveillance, it’s not necessarily talking about “what happened today” but it’s really “is there a bad actor in the market.” You want to look for suspicious transactions. But if you want to explain what happened, you need the full data suite of when the client first called the broker, or when the proprietary trader first placed an order into the market, at what time? What was the order? What were the conditions of the order? The entire suite. Books and records are required, so that if CFTC or the SEC wanted to look at a swap transaction, and the price formation of it, the participants are supposed to keep detailed records. Well, records systems for recording, change over time. It’s difficult for some people to access their own records. I put out special calls with firms asking for three plus years, five out of five years (in some instances), worth of data, and they would come back and say, “Our record-keeping systems were updated, we can’t inflate our data anymore,” or “it’s not accessible because we’ve made software changes.” It’s problematic, right? If it’s held in a central repository, it’s not problematic, you don’t have to rely on the participant. My own view of algorithmic trading:  if you recall, CFTC changed its rules that essentially said to call for the algorithms that are being placed into the market, a subpoena has to be issued. Where at one point, I thought of them as books and records and I would call for that algorithm using a special call authority, which has the ability to reach books and records. And the idea is that the market participants would say, “Well, we don’t really know what we placed, we placed so many different orders” and the algorithms are self-modifying. So, you go “That’s really great,” except if you have an event and you say, “What happened, which order went, which algorithm?” What did that algorithm do and how is it structured? If they’re not collected, then you can’t see if there was a flaw in the program itself. So that’s a long-winded explanation.

Asif

No, thank you. Thank you. So, we have talked about data. And to me, data is critical in a lot of things that we are doing now, Matthew. Let’s switch a little bit on to something different, which is about AI potential. So, what about AI (all of a sudden?) AI has been around for many years but right now, there has been a lot of buzz around AI. It is used today. What about its use in the future of surveillance purposes, prosecutions, and by regulated firms and internals, compliance monitoring use. How can AI be something that you can use in that sense?

Matthew 

Well, I think AI is unbelievably important. And I’m going to step back with a little story. So, a number of years ago, when I was at the CFTC the opportunity arose for me to train “Watson,” the IBM AI, in the detection of market manipulation, (benchmark manipulation, particularly) and it didn’t come to pass. It’s my biggest disappointment, my biggest regret of being at the commission that I didn’t get to use this tool. And the reason is that you only have to teach the machine once. AI is rules-based, all surveillance is rules-based, many of my staff would say to me, “But our problems are not that complex. You don’t need machine learning necessarily, or a highly intelligent machine to do the work.”  Benchmark manipulation has only a very limited number of rules. (According to me anyway) and I thought, “What a disappointment” because the machine never forgets. You can bring in all your case history, you can bring in all your successes and your failures, right? You bring in the data. If you have a fairly complete data set, you can make a highly compelling story for why someone did or did not violate the rules. So, my own view is that AI is a tool that equips compliance staff with the ability to review their own data very quickly. All right, looking for, pointing out spoofing. In other disruptive trading, by rule and benchmark manipulation, attack on settlement prices, the trade to enhance what I call a “benefiting position.” So AI, I think, is the tool of the future. Rather than trying to do it through so-called “piecemeal,” one thing after the other. The world is too complex, you should be able to get a report at the end of the day, no later than the next day next morning, about the behaviors of trading staff, based on the rules of what comprises violations.

Asif

Yeah. Matthew, just now when you were talking about Watson, one thing that comes into my mind is, I think it’s about timing. I remember very vividly, I used to be in New York, and now I’m in San Francisco. I used to be in New York in 2008 and was very much involved back then on trading technology at the time. What I saw post-2008, the whole financial crisis was that technology or “cloud” became a must-have, not a “nice to have.” All the complications prior to the idea of, “Hey, can I be on Cloud?” Can I use technology? That just went away, because of the complexity and the velocity that came out, and some of the struggles that came out, it became a must-have. Fast forward to 2022 and 2023. What I’m seeing right now is that, again, with the same narrative of complexity and velocity, going at another level, the use of technology is becoming a must-have. Reg Tech, all those things, again, (from a regulatory perspective) and, yes, you can do things (to your point), you can do things manually. But when you have something as powerful as technology right now, that can really enable you and simplify, why not? And that’s what I’m seeing in most cases. So, it’s really about enabling, it’s about risk mitigation. It’s about repeatability that I’m seeing now where I see a lot of that and less of, “hey, let’s not do that, let’s move away from that very manual process,” (if you will), to using technology and AI in that sense. Talking about AI, you know, these days, there is a lot of buzz. There are a lot of different kinds of AI but obviously generative AI, Chad GPT and others. What is your view on that? Do you have a view on that and how useful or not useful? What’s your point of view on that?

Matthew

Okay, so on generative AI, clearly it is -call it a shortcut mechanism-for people today. I call it just a shortcut. Ultimately, I think that some of the doomsday scenarios about generative AI are probably right. And the doomsday scenario is: how many hundreds of millions of jobs will be replaced? You know, do you need accountants? Do you need lawyers? Do you need any number of rules-based jobs (right at the moment) that are really about typing. One of our journalists that writes about that certain thing it doesn’t do and you have to, (if you haven’t already) signed on to chat GPT and played with it. Just ask questions and go. It’s a fascinating walk into the beginning of a technology. We’re not seeing the most advanced form of AI and can’t really do data analytics with it. Yet, there will be I’m going to guess, three, five years from now, Analytic based AI that’s accessible to everybody. And if it has the data set, (including extremely large datasets and generative AI capacities) it will be able to do analysis for you based on rules. But I played with, with AI and asked him to do an analysis of different commodities, I got very plain vanilla answers that certainly made sense with the information that was available to others time. That was in the press, I have also asked it to do more complex, original, call it “thinking” to generate things that haven’t been written about before,  using very arcane, limited sources, literary sources. So I believe that it’s going to be part and parcel of virtually everything that we do going forward as a tool to enhance and speed human beings in doing some tasks. And there is a danger that we’ll replace people with those tasks entirely one day, but not today. Not today. It’s still pretty primitive.

Asif

Yeah. And my view there, Matthew, very quickly on AI, is that chat GPT and others, Yes absolutely, it’s fun to play and get the answers. Sometimes it answers the right way and sometimes it does not. When you bring it to you in a business scenario, you need to be very careful, because the information you’re getting will get better, but it’s not there yet is what my view is. And because one thing that we take a lot of pride in at Compliance AI, for example, just to give you one example from our side, is that yeah, our name says compliance.ai  AI is, as you know, our bread and butter in that sense, but we have something that we call “expert in the loop”, which is to make sure and ensure that the obligation that is being extracted, the information that we are doing, that there is a human verifying and validating, the answers that our customers are seeing. What I’m seeing right now with Chat GPT is that, yes, it’s learning from some of the mistakes, but if you take it to the business case, there are still a lot of gaps in there. And you must be extremely careful in those scenarios, in my opinion,

Matthew

I agree with you. And so, I’m going to jump a little bit I got asked a question through LinkedIn, that if you don’t mind, I’m just going to ask myself. I’m going to read it, the question is, “Looking beyond detection, I’m curious how regulators will prove the manipulation they find in an AI’s operation is largely a black box. will humans be able to explain to the jury the complex connections that AI finds in the markets and products, which connections presumably current human program data screens have missed? And I think the answer to that is, in the end, what does AI look like 25 years from now? It’s entirely possible that judicial findings will be supported nearly completely by AI as long as it’s rules-based. The thing that AI cannot do today, and I doubt will do in the foreseeable future, and I don’t know what that means is it 5,10, 20, 30 years some significant number of years, it doesn’t have the ability to have insight. And, to make leaps in logic to a completely new conclusion. So, you know, making new precedents within law is unlikely to be supported immediately by AI. But I put an asterisk next to that. And the asterisk is this: The AI was used to challenge the Go masters, and go is the most complex game that human beings play on aboard the AI is made in one master’s game, what looked like a blunder, a rookie mistake. And it won the game. And that move was pivotal to changing human beings understanding of how go could be played. And it was a radical move, as opposed to a blunder. And that’s what I’ve read. So I’m using the idea that it can have, potential insight. So, I would not say that Chat GPT currently is geared to making novel expressions and, and having an evolutionary or progressive thought, it’s still a machine that’s rules-based and gives based on input and rules. So I think that to answer my friend’s question, in the short term, you need the depositions, you need the traditional, acts of attorneys, but I don’t know that this will always be the case with the machine and I don’t know that the machine can explain exactly the logic that was applied, and what the case precedents and what the data said that can’t be put out as evidence. Okay.

Asif

So, as we are concluding towards the end, I would if anyone has any questions, please put them in the chat. We will love to get some questions as we are coming to an end. But before we end, one last question that I want to ask, Matthew will the recent enforcement actions (in all likelihood) have a major impact on the internal compliance team and also market participants?

Matthew

Absolutely. I read with great interest, a recent case where CFTC fined, BBL for supervisory violations where effectively they said that the firm did not provide sufficient training in disruptive trading. Now what they did, (what the case is about) is placing large orders essentially during the settlement period of a spread that was sufficient to move the market. Normally I would care that’s a manipulative practice, but there was no description of the intent behind the trade or the resolution of the trade. He only said that this was a disruptive act, but they were holding the firm responsible for how we trained, and that to me, brokered up. And an understanding of if disruption is by placing market size or off-market size orders into the market. That means that the traders have to be trained in assessing market liquidity and have to have some kind of benchmark for understanding what an acceptable move in the instrument in the commodity is. In this case, the spread. And I don’t know how that’s done, but I understand it needs to be in the compliance materials. So, if I believe that, the generative AI, when queried about that case, and this is something that I believe is within the capacity of Compliance AI’s program, is to point out what changes must be made in the training documents that deal with this enforcement action. So beyond mentioning that disruptive trading practices are prohibited, actually outlining what those are and providing that training, I think is very significant. It also means that enforcement agencies are going to get into the weeds about what was and what was not done. Even if they don’t go after the more significant issues of price manipulation, or schemes to move value. So, I think that case was really important. When I think about my career, (I know we’ve talked about this and I’ve said this to many people), over time, I was always interested in rule changes, and in the obligations that I had on the desk. And this is one of those instances where I think it’s not the supervisory aspect. That’s important, right? That’s always important. It’s always out there. But it is the level of detail that is now being impacted. And I think that that could be picked up by the program, right? Compliance people are interested in what does this mean for me. When enforcements or rules come out.

Asif

Awesome. Ronjini? Do you see any questions?

Ronjini Joshua

no questions, we had a nice comment about chess computers, similarly eliminating humans with a brilliant new strategy. I think that was the comment about speaking to the AI part of the conversation you guys are having.

Asif

Awesome, awesome. So, if you can go to the next slide very quickly. And the way I put ourselves, is that we provide, frankly, compliance orchestration to really enable professionals in the banking, financial services, and insurance to monitor track reports, and react to record changes and obligations in real-time, using the technology. And some of the things that I say which is a core to our capabilities, which is big for us is in as you, as you see these pillars, is noise reduction of getting those 2000 Plus sources 200 different types of documents, auto-translation from 20 plus different languages, to have an impact analysis of faster obligation 75% Faster obligation to do automated change management, which is key, you know, about I think our numbers are 30 plus percent reduce operating costs. And to the fourth and fifth pillar, is monitoring enforcement action, as well as audit reports of fewer compliance, penalties incurred, and having a complete audit trail. Those are some of the things that we are very proud of actually delivering to our customers. And something new that we recently added to our solution as of last year, brings your own content, which is really bringing your own any kind of content policies, anything that you can upload to compliance AI and in real-time and almost real-time can check if they are, you know, compliant to what you’re doing or maybe you’re going through an m&a or you’re doing certain activities to see in future what it means in a certain state in a certain jurisdiction, how you do that. If you please, reach out to us if you guys are looking to implement, if you have not already, Reg Tech, as I said earlier, and I think Matthew said it also that the complexity and the velocity absolutely is there. And what that really means is that the time for implementing and executing and mixing, enabling technology with that would be a big thing. So, there are some email addresses here, and our website is here. Thank you very much, Matthew, for your time. We really appreciate it on behalf of everybody at Compliance AI, and thank you, to everyone who participated, there will be a recording if you were not able to listen to this. And if you think this would be helpful to your colleagues or somebody else. We will be having a recording also soon on that. So, thank you very much. Ronjini. Back to you.

Ronjini Joshua

Thank you. Yes, like Asif mentioned, we are recording this webinar. It’ll be available early to mid-next week. And we’ll be sharing it with all the attendees if you miss part of this conversation or like you said month wants to share it or pass it along. Thank you so much. And we look forward to joining us next time in our May webinar. Thank you so much. Thank you, bye.

Regulatory Intelligence

The Potential of AI

X