Johny Fernandez: And welcome back to Market Movers: Midday Edition. So prediction markets are seeing a surge in trading ahead of this year’s midterm elections, and with a busy stretch across the sports world right now, bettors have plenty to wager on. And meanwhile, the debate over privacy is heating back up. Now, in crypto, privacy-focused tokens like Zcash have been rallying strong. And in AI, the fight centers on how companies collect and use people’s data, and whether that crosses the line of privacy.
So joining me to discuss this and more is Kelly Cherniwchan, founder and CEO of Chata Technologies. Kelly, thank you so much for joining us today.
Kelly Cherniwchan: Thanks for having me, Johny.
Johny Fernandez: So let’s go ahead and start with prediction markets, because obviously prediction markets are very popular right now, especially when it comes to things like the elections, and even cable news channels are citing them for election forecasts. And they’re becoming a bigger signal of the overall economy. So what’s the outlook for using them to hedge or protect against risk?
Kelly Cherniwchan: Yeah, I think the majority of prediction markets now are focused on sports betting. I mean, that’s what you see in the media. Where we believe prediction markets are going to go is around, like you mentioned, hedging, right? All businesses are affected by economics, commodities, everything else like that. And typically, they have to go to large brokers in order to, you know, work with exchanges to be able to hedge their inputs and their outputs as well.
Prediction markets now allow medium businesses, even small businesses, to take those contracts on canola prices or diesel prices or whatever it is, in order to smooth out the volatility of their inputs and smooth out the volatility of their outputs as well, so that they can just focus on their core business.
Johny Fernandez: So, Kelly, there’s been a lot of pushback against prediction markets also, and not just from states. We know that there were a couple of studies this week that found that most people use them for gambling. So how do they prove their real value? I think that’s the first question, as a source of real-time data for financial markets.
Kelly Cherniwchan: Yeah, I know, it’s a great, great question. So as volumes increase, especially in the economics contracts and the commodities contracts, and you’re seeing growth there, even in perpetuals as well, it’s less of a gambling approach than it is in sports. This is being utilized for price discovery, right? So you have traditional market pricing. You have the prediction markets. There’s some research out now around how, in certain cases, CPI is better predicted by prediction markets than by the financial markets. So as the historical data builds up about how accurate these prediction markets are, you will see a large increase in the amount of usage in the area.
Johny Fernandez: So I want to continue this conversation, because we also know that states are fighting the CFTC over who gets to regulate prediction markets. That’s been a big topic of conversation. So will the Supreme Court end up deciding who’s in charge? And once the legal fight starts to settle down, how soon could investors start relying on them?
Kelly Cherniwchan: Very, very good question. I don’t pretend to know what the Supreme Court is going to do. But I think if we approach this logically, what there is going to be is a bifurcation in the sports arena, because state regulation of commodity and economic indicator-based prediction markets just doesn’t really make any sense. That should be squarely in the domain of the CFTC.
When you start even talking about game and futures levels, who’s going to win the NBA championship, who’s going to win the NFL, there’s a case for hedging in those instances, for restaurant groups, for anybody whose business is tied to that. That still could remain under the CFTC. Now, the prop betting markets, player statistics for a game, will Patrick Mahomes throw over 200 yards? That is bordering on the level of gambling. And therefore, there could be a bifurcation of prediction markets: which are state-regulated, and which are actually regulated by the CFTC. So we’ll have to see. But I just look at where the logical separation is between those markets, between what can be used for hedging and what’s gambling.
Johny Fernandez: All right. So I want to turn our conversation now to AI, because obviously AI is top of mind. We just saw Meta hit with a $25 billion penalty for misleading users on how it is using their data. So where does the debate stand over using consumer data to train AI models?
Kelly Cherniwchan: Yeah. So, training AI models: a lot of the large vendors are coming out with policies that they’re not going to utilize the data of their customers to train the weights of the AI models. And this is in their terms and conditions. But there is a big difference between training the models with that knowledge and still storing that knowledge and potentially accessing that knowledge. So that’s a big differentiation.
So when you think about privacy, think about a sell-side person at a large bank investigating client trading activity, because they’re looking at block trades to do between different clients that are mutually beneficial. That information cannot be going to the major AI vendors, stored in QA, stored in anything. And it is happening all of the time now. Companies, as they’re using Anthropic, as they’re using OpenAI agents, are literally disclosing everything. Maybe it doesn’t get used to train the weights, but it is still being stored. It’s confidential information that’s being stored that, unfortunately, could be utilized in certain cases for the benefit of those large vendors.
Johny Fernandez: And so we’re also hearing more and more about agentic AI, where AI assistants act on your behalf. So what are the privacy risks when consumers give these agents access to everything?
Kelly Cherniwchan: Yes. Yeah. Access to everything is worrisome, for sure. And there were some articles out today as well about how some of the large US banks are noticing activity with agents that’s actually not authorized by the individual account holders, or perceived not to be authorized. When you give an agent access to your API keys and all of your credentials, it does have the ability to go in and do whatever it wants to do with that. And remember, the core technology of these APIs is still based on large language models. No matter how many guardrails you put in place, they still have hallucinations, right? And some of these hallucinations and the outputs are quite scary, for sure. So that’s critical.
The second piece is people using it in financial markets to execute on trading strategies, or look for trading strategies, or investigate data to come up with new trade ideas. All of those interactions with the agents are being sent to the company, which is storing all of those interactions on its side. So privacy-wise, banks, hedge funds, prop firms all have to be very careful when utilizing these systems, so that they’re not giving up all of their core secrets.
Johny Fernandez: And so let me ask you this. I know this conversation is happening a lot when it comes to banks and payment companies. So what are the biggest privacy worries for banks and payment companies when it comes to agentic AI?
Kelly Cherniwchan: Yeah, just knowing all of the different flows of money and to whom they’re going, right? So if someone’s interacting with an agent and they say, how many trades has Kelly Inc. made on this particular gold contract over $10 million? That data is now stored. That is proprietary data. So myself, as a potential trading client of a broker, I don’t want that information going out to third parties where there is no strict confidentiality between them, right?
So I think what you’re going to see, unfortunately, is probably more and more litigation around this area of privacy: wealth clients having information disclosed, you know, traders having information disclosed, banks. And it’s all going and being stored, and can be reviewed by quality assurance people, as an example. So there’s a lot of regulation that I think still needs to be put around this topic.
Johny Fernandez: No, it’s definitely a topic that’s always ongoing. But now let’s talk a little bit more about Chata. Your technology lets users search company data just by asking questions in plain English, without exposing how that data is being stored behind the scenes. Today’s leading AI models want to take more and more data. So how do you give users detailed, useful answers while keeping strict security controls and not storing any of their data?
Kelly Cherniwchan: Yeah. So, that example I gave you about the sell-side person looking at Kelly Inc. trading gold contract XYZ over X amount. What the technology does is, within your secure environment, before that ever gets sent to an agent or an LLM or anything like that, that entire query gets obfuscated. So when that query leaves, it doesn’t know that it’s Kelly Inc. It doesn’t even know the contract, if that’s sensitive data, and even the amount as well. So to those large language model systems or agents, it doesn’t know anything.
What comes back into the environment, where we build these database queries, is a proprietary language that we built, where we just get instructions back. Then we build that database query to get you that information, to find out, you know, if you’re looking at counterparty trading activity, if you’re looking for very specific targets out of the client list, if you’re looking for very specific prediction market contracts. All of that happens within your own environment, and none of that proprietary information gets disclosed in that natural language interaction outside of your environment.
Johny Fernandez: Well, it’s definitely interesting to see how AI continues to develop. But Kelly, thank you so much for joining us. I definitely appreciate you and your time.
Kelly Cherniwchan: Sounds good. Thanks, Johny.
Johny Fernandez: No problem.
