One fraud you might try is:

1. Find a house that costs $1 million.
2. Go to a bank and ask to borrow $800,000 to buy the house.
3. The bank gives you a mortgage loan, you put in $200,000 of your own money, and you buy the house for $1 million.
4. List the house for sale for $900,000.
5. Quickly find a buyer (because you are selling the house at a discount).
6. Say to the buyer “I want an all-cash deal.”
7. “No problem,” says the buyer, “I have no mortgage contingency and can close quickly.”
8. “No I mean I need a lot of $20 bills in sacks.”
9. “Uh okay I guess,” says the buyer, and hands you cash in sacks.
10. You hand the buyer the keys. “Pleasure doing business with you!”
11. You don’t pay back the bank and skip town with your $900,000, for a $700,000 profit.

I am sure that someone has done this successfully at some point in human history! It is quite hard to do in the United States in 2025. Most of the apparatus of modern US real estate is about preventing this. When you buy a house, there is a closing, a ceremonial occasion whose purpose is (1) to make sure that the seller’s mortgage gets paid off before the seller sees a dime and (2) to assure the buyer that she actually owns the house. Claims on the house — like a mortgage — are publicly recorded, so the buyer (and her lawyer) can see that there’s a mortgage on the house that needs to be paid back. But I suppose there are people out there who (1) have $900,000 in cash and (2) don’t know about real estate closings. [1] Maybe you could get them to give you sacks of cash for the keys, and then skip town. It does not seem especially worth trying.

What about doing it at one remove:

1. Find a house that costs $1 million, and someone who has $200,000 and wants to buy it.
2. Lend the buyer $800,000 against the house, which he buys for $1 million, giving you a mortgage on the house.
3. Go to a bank and say “hi I am in the real estate lending business, I’ve got this mortgage here, would you lend me some money against it?”
4. “Sure,” says the bank, and gives you $500,000, secured by your mortgage.
5. Go back to the homeowner and say “hey I’ll buy that house back for $1.05 million.”
6. “Sure I guess,” says the homeowner, who is perhaps your buddy and was always planning to sell you the house.
7. You give him $1.05 million, he repays your $800,000 mortgage and keeps $250,000 (a $50,000 profit, given that he put up $200,000 in Step 2).
8. Now you own the house. You are out, net, $550,000 of cash (you made a mortgage in Step 2 that was repaid in Step 7; you paid $1.05 million in Step 7; you got $500,000 from the bank in Step 4).
9. Find someone to buy the house for $900,000. This is easy: You’re selling the house at a discount, and there’s no mortgage on the house (it was repaid in Step 7), so there are no liens and no reason for the buyer to be suspicious.
10. Now you are up $350,000 in cash ($900,000 from Step 9 minus $550,000 from Step 8), though you owe the bank $500,000, so this is not a good trade.
11. Unless you skip town!

You can find variations. (In Step 5, instead of you buying the house, perhaps the homeowner can sell it to someone else and pay off your mortgage. Or perhaps you can sell the mortgage loan itself to another investor.) You have borrowed money from the bank, secured by your loan secured by the house, and then you get your loan paid back (or sell it) without paying back the bank.

The insight here is that there is a really well-oiled process for liens on houses, and if you try to sell a house without paying off your mortgage someone will probably stop you. But is the second-order process — for liens on mortgages — as well-oiled? If you try to pay off a real-estate lender without paying off the bank that financed the real-estate lender, will anybody stop you?

I confess that this is not a question I have generally thought much about. I suppose I would have guessed “yes, of course, the real-estate lender is a company that makes lots of loans and has some big credit facility with a bank, the bank probably keeps track of this stuff pretty carefully, it files all the right financing statements in all the right places, probably there is some mechanism to make sure the money goes where it is supposed to.” I am not sure what I thought the mechanism was but, you know, surely this comes up a lot, surely they’ve got it figured out.

But we talked last month about alleged double-pledging of collateral in connection with the failures of Tricolor Holdings and First Brands Group Holdings, two car-related companies that might have sold, not cars or car parts but rather amounts owed on cars and car parts, more than once. I wrote:

The point is that it is hard to sell the same physical product more than once, but it is easier to sell the same abstract financial product (like a loan) more than once, and in a world where products are normally sold on credit, you can kind of sell anything twice. 

Similarly, perhaps, it is hard to sell a house without paying off the debt on the house, but perhaps easier to sell the debt on the house without paying off the debt on the debt on the house. Maybe? It goes without saying that everything in this column is extremely far from being legal advice.

Anyway:

Shares of Zions Bancorp and Western Alliance Bancorp plunged Thursday after the companies said they were victims of fraud on loans to funds that invest in distressed commercial mortgages. The disclosures sent the KBW Bank Index to its worst day since April’s tariff tantrum.

They followed a warning from JPMorgan Chase & Co. CEO Jamie Dimon about “cockroaches” in the credit market during the bank’s earnings conference call on Tuesday. He was referring to the implosions of auto lender Tricolor Holdings and car-parts supplier First Brands Group. 

Bloomberg’s Yizhu Wang has more on the alleged fraud:

Zions said in a lawsuit that [its subsidiary] California Bank & Trust is owed the money from two investment funds tied to Andrew Stupin and Gerald Marcil, among other parties.

California Bank & Trust provided two revolving credit facilities to the borrowers in 2016 and 2017 totaling more than $60 million, to finance their purchase of distressed commercial mortgage loans, according to the lawsuit. The terms gave the bank “first-priority, perfected security interest” in all collateral, including each mortgage loan purchased by the investment funds.

But after an investigation, the lender found that many of the notes and underlying properties were transferred to other entities, and that those properties have already been foreclosed on or were about to be, according to the lawsuit.

“My clients vehemently deny all the allegations of wrongdoing,” Stupin and Marcil’s attorney, Brandon Tran, wrote in an emailed statement. “These claims are unfounded and misrepresent the facts. We are confident that once all the evidence is presented, our clients will be fully vindicated.”

And:

Western Alliance also lent money to the same investor group, for them to originate or buy mortgage loans, according to the bank’s lawsuit in August. The outstanding balance of that loan is $98.6 million.

Western Alliance found that the collateral was supposed to be backed by a first-priority lien, but that wasn’t the case. It alleged that the borrower created fake title policies by omitting the senior liens.

And Wang had another story last week about ties between the Cantor investors and a real estate investment firm that got some of the loans at issue here.

Here are the announcements from Zions and Western Alliance. And here are the lawsuits that Zions and Western Alliance brought against the Stupin-and-Marcil investors (who invested through entities called Cantor Group II, IV and V). The allegations are different, but they rhyme. Zions says that the Cantor investors borrowed from CB&T, secured by some mortgages that they had on commercial properties, and then sold those mortgages without telling CB&T that its collateral was gone:

CB&T has determined that, for each of the underlying collateral real properties for the notes pledged—or intended to be pledged—as collateral, CB&T’s security interest was not a first-priority lien, as expressly required by the Security Agreements. To the contrary, the deeds of trust in favor of Cantor II or Cantor IV (and, by extension, CB&T through its security interest in all collateral) were reassigned, diverted to other Cantor-affiliated entities, and/or never properly recorded. In numerous cases, the notes have been transferred to other entities, the underlying properties have already been foreclosed upon or are currently the subject of foreclosure proceedings. …

And: Many of the new senior lienholders who displaced CB&T’s security interests were, in fact, entities owned or controlled by the same individuals behind Cantor II, Cantor IV and the guarantors—including Cantor V. In effect, CB&T’s losses became Defendants’ gains. Acting through a web of affiliated companies, the borrowers and guarantors orchestrated a scheme that enriched themselves while stripping CB&T of its collateral, all while keeping the bank in the dark for years about the true status of its security interests and the absence of any collateral protection for the more than $60 million plus in credit it had extended. 

While Western Alliance says that the Cantor investors doctored title documents to trick Western Alliance into lending money against first-lien mortgages that weren’t:

Because a pledged loan is “Eligible” only if WAB’s lien on the underlying real property is valid, first-priority, and perfected, Borrower [Cantor] began providing WAB with doctored title policies purporting to show that WAB’s lien on the real property underlying certain pledged loans was first priority. …

On October 15, 2024, Borrower’s employee Jaspreet Singh Sethi provided to WAB a purported Stewart Title Guaranty Company Title Policy No. M-2923-14382 dated October 1, 2024 related to Collateral Loan No. 26, which involves property located at 2522 S. Grove Ave., Ontario, CA with property owner Conejo Riverside Group LLC.

The purported title policy shows WAB’s lien on the real property underlying the property as first priority. ...

However, a title search run by WAB after entering into the First Amendment to the BLSA shows a deed of trust for Preferred Bank for $10,400,000 dated July 11, 2017 recorded July 17, 2017 in the first priority position. 

The suggestion here is:

1. There is a public record of mortgages, so anyone can search the record and see if there is already a mortgage on a particular property.
2. Therefore it is basically impossible to “double-pledge” a property: If you have already borrowed $800,000 against your $1 million property, and you go to another lender to borrow another $800,000, the second lender will do a title search, see the existing mortgage, and say no.
3. But what if you have a buddy who is a mortgage lender? What if you go to him and ask to borrow another $800,000 and he says yes? Then you have borrowed $1.6 million against your $1 million property.
4. Then your buddy can go to his bank and borrow, say, $500,000, secured by his $800,000 mortgage on your property.
5. His bank might not check the title: The bank might say “hey this is a first-lien mortgage right,” and your buddy will be like “oh sure here’s the title report,” and the bank might not do its own search to check. Why would it? Your buddy has skin in the game; he has incentives to make sure that he has a good lien on the property, so that he can get paid back. Or so the bank thinks.
6. You and your buddy have, between you, extracted $800,000 from the first lender and $500,000 from his bank, for a total of $1.3 million against your $1 million house.
7. You skip town together.

It’s hard to double-pledge a commercial property, but it’s easier to double-pledge a commercial property loan.

A framework that I think about a lot is Nick Whitaker and Zachary Mazlish’s 2024 essay “Why prediction markets aren’t popular.” Their answer is that “there is little natural demand for prediction market contracts,” because they are “not a natural savings device” and “not a natural gambling device.” Traditional financial markets “attract money from pensions, 401(k)s, bank deposits, or brokerage accounts,” as well as gamblers “who enter markets for thrills,” but prediction markets are not a way to build wealth and not a fun gamble. Traditional markets have a lot of investors who are “not attempting to beat the market by correcting pricing errors,” so if you are attempting to do that, you will be able to make much money trading against them. Prediction markets lack that supply of price-insensitive investors, so there is not much reward for correcting prices.

If prediction markets are going to be big and influential, they have to solve this problem, by becoming fun gambling markets or by becoming useful savings vehicles or both. [2]  The obvious obvious obvious obvious solution is of course to become fun gambling markets: The structure of a prediction market is “you can buy contracts that pay out if your predictions about real-world events come true,” football games are real-world events, and betting on football games is fun. Problem solved. We talk about this all the time: For some bizarre and accidental combination of reasons, prediction markets have become the legally favored way to do sports gambling in the US, and they have jumped enthusiastically into that business. 

What about savings? The reason that people put their 401(k)s in stock market index funds is that stocks generally go up. A share of stock represents partial ownership of the future profits of a company; an index fund represents partial ownership of the future profits of all the companies; the economy generally grows over time, so the profits of all the companies should grow over time too. So just lazily putting your money into the stock market should get you a nice return (the “equity risk premium”), so people do that, which means that other people can make a living by making stock prices more correct.

Is there any analogue in prediction markets? No? The stock market is broadly positive-sum, as stocks mostly go up with economic growth, but prediction markets are obviously and necessarily zero-sum: Every dollar I make comes at your expense, so there is no way for all the investors to put billions of dollars to work in prediction markets and all earn a positive return. [3] There is no equivalent of the equity risk premium in prediction markets, no equivalent of index funds, no simple trade that everyone should do with positive expected value for all of them. 

Although my Bloomberg Opinion colleagues Carolyn Silverman and Tim O’Brien found something pretty close:

We examined over 300 markets on Polymarket pertaining to actions directly initiated by Trump this year, between Inauguration Day and Sept. 30th. That analysis included everything from tariffs to whether he’d fire a Cabinet official to what executive orders he’d sign. Bettors even wagered on his golf game and the duration of Elon Musk’s White House tenure. On the day each market opened, bettors assigned an average probability of 34% to those events occurring within a specified timeframe. But only 28% actually happened.

While a sophisticated forecaster could have devised a complex betting scheme to exploit this miscalibration, a simple rule would’ve also made you money: Always bet “no.” Betting against Trump taking any action would’ve translated to a 12% return — similar to the S&P 500 index’s gain over the same period. Consistently betting “yes” would have cost you 20% of your money. (We aren’t offering investment advice; we’re just sharing the data.)

Ahahahahaha. There is an intuition for this, isn’t there? Like:

1. The a priori probability of President Donald Trump doing any arbitrary insane thing is, like, 20%.
2. If you are logging on to Polymarket to bet on Donald Trump doing any particular insane thing, it’s because you have some idiosyncratic reason to think he will do it: You want him to do it, or you are hedging some bad consequence to you of him doing it, or you have an inside tip from someone in his administration that he’ll do it, or he publicly announced that he’ll do it.
3. Therefore the contracts are systematically priced too high: Everyone is coming to Polymarket to bet on their pet insane things happening, but he’s only got so many hours in the day so only so many insane things will actually happen.
4. Therefore if you systematically fade everything you can reliably make money.

It is not quite index funds: It’s still zero-sum, it wouldn’t work if everybody did it, and it is not exactly an investment in the long-term growth of the US economy. And yet somehow it feels like a … sensible medium-term macroeconomic bet? “Everything will be crazy for the next four years, but each particular thing will be slightly less crazy than it looks at first”? Apparently you can make a steady index-fund-like return from that bet.

Elsewhere!

CME Group Inc. is planning to debut financial contracts tied to both sports games and economic indicators by the end of the year, according to people familiar with the matter.

The products would bring the Chicago-based CME into more direct competition with Kalshi and Polymarket, the two platforms for so-called prediction markets that have been making rapid moves into the financial mainstream with the support of Donald Trump Jr., who is an adviser to both companies.

CME is preparing to release its new contracts to the public through futures commission merchants, including the one it is setting up with FanDuel, a part of Flutter Entertainment. But CME could also offer them on other similar platforms, opening it up to traditional retail investing brokerages, according to the people, who declined to be identified because the plans are still private.

Sure!

I guess my thesis would be that, 10 years ago, if you were an optimist about the capabilities of artificial intelligence, there were a few broad categories of stuff that you might be working on:

1. Natural language stuff: Responding to customer service queries, translating from one language to another, etc.
2. Visual stuff: Identifying traffic lights in photographs, etc.
3. Ad stuff: Using data to predict what advertisements to serve with what websites, and how much to charge for them.
4. Financial stuff: Using data to predict what stocks will go up.
5. Astrophysics stuff: Finding black holes or something, I don’t know, but the link between astrophysics and all sorts of data techniques seems weirdly robust.

These are all broadly similar endeavors. With each of them, you can take some training data (text, pictures, market states), mask something (the next word, the next pixel, the price of the stock the next day), and train the model to predict the masked data point. Do that a bajillion times and the model might get really good and able to do it out of sample. 

And so, as the capabilities of AI really did advance, people moved back and forth among these endeavors. Renaissance Technologies is a hedge fund that broadly seems to use AI to predict stock prices; many of its leaders came from the world of computational linguistics. DeepSeek is an important builder of AI chat models; its leader came from a hedge fund. Astrophysicists are everywhere.

But it seems to me that, in 2025, the visual and natural-language people have moved ahead of the financial people. I don’t think it was obvious in 2015 that this would be the case — I think that in 2015 it might have looked like the practical thing to do with AI skills was going to a hedge fund — but now several generative AI labs have come out of nowhere to achieve huge valuations and everyone who works on a chatbot is a millionaire. They have made popular consumer products and upended businesses and deployed vast resources and turned their leaders into celebrities. They have published papers and open-sourced some of their work. There are lots of chatbots available for $20 a month that will converse with you in natural language, produce well-researched reports for you, and draw you anything you want in the style of anyone you want. There is not a $20 model that tells you what stocks will go up.

And so we are in a slightly odd world where people ask the chatbot models what stocks will go up. I once wrote that this approach — using a general-purpose natural-language chatbot to predict stock prices — seems “sort of insane and wasteful and indirect”:

Yet also funny and charming? It is an approach to solving the problem by first solving a much harder and more general problem: Instead of “go through a ton of data to see what signals predict whether a stock goes up,” it’s “construct a robot that convincingly mimics human consciousness, and then train that robot to mimic the consciousness of a particular human who is good at picking stocks, and then give the robot some basic data about a stock, and then ask the robot to predict whether the human would predict that the stock will go up.”

There are versions of this for retail investors: Your retail brokerage might have a chatbot that gives you portfolio advice, or you might just ask ChatGPT yourself. But meanwhile all the big banks and hedge funds are also experimenting with generative AI, not so much in the sense of “we trained a neural net to find what stocks will go up” but rather in the sense of “we got a ChatGPT license for everyone so they can ask it for research summaries and help with Python.” Bloomberg’s Hema Parmar reports:

When Viking Global Investors rolled out its latest in-house artificial intelligence chatbot, traders and dealmakers wasted no time putting it to work.

Teams across the $53 billion firm sent three questions per working minute to “VikingGPT 2.0” in August — a 400% jump over the same period last year. The proprietary generative AI model is shouldering grunt work, helping staff bat around trade ideas and speeding up operations, according to an investor letter seen by Bloomberg. In one instance, Deputy Chief Investment Officer Previn Mankodi used it as a “skeptical avatar” to challenge an investment thesis.

Here’s what it’s not doing: making wagers on its own.

“Lest you worry we are ceding key decisions to the machines — quite the contrary,” billionaire co-founder Andreas Halvorsen wrote in the investor letter. “We have observed that the closer you get to needing to make a judgment call, the less useful AI is.” ...

The letter sheds light on the ways top money managers are using generative AI, a technology some view as a potential game-changer and others say is overhyped.

[Last] week, Citadel founder Ken Griffin said the technology isn’t helping hedge funds create market-beating returns and hasn’t yet meaningfully impacted the industry.

Counterpoint: From the outside, it sure seems like AI has helped Renaissance create market-beating returns for many years. But I know what he means. Asking a chatbot to pick stocks doesn’t work yet.

I guess there are two theories of an investment bank’s headquarters. One theory is that it is an enduring monument to the bank and the visionary chief executive officer who commissioned the headquarters. Here’s Jamie Dimon on JPMorgan Chase’s fancy new Norman Foster headquarters in Manhattan:

“This building is a labor of love,” Dimon said, gesturing at the steel structure’s dark-bronze facade. “It is an embodiment that represents our company and where our people are going to work.”

He knows he’s not going to be around the building forever. But the fortress, visible from Brooklyn, Queens and even lower Manhattan, if you look just right, will remain long after he retires. 

“We move paper, that’s what we do,” Dimon said. “We move paper around. This is a permanent thing.”

The other theory is that it is a building with desks and computers where the bankers sit and move the paper. Of course both theories are true; the headquarters can both stand as an eternal monument and also have enough desks. But in practice the two theories are in tension:

Some investment bankers are finding that the building’s tall floors and large common areas mean their actual working spaces are more cramped than those at the old digs across the street at 383 Madison Avenue, people who work at the bank said. 

You can fit more desks if you don’t have a four-story golden statue of Jamie Dimon in the lobby but what’s the fun of that.

BDCs, Private Credit’s Most Popular Funds, Are Drawing Scrutiny. Sold Out in India, Panic in London: How the Silver Market Broke. AWS Says Cloud Service Has Recovered After Widespread Outage. Basis Trade Helps Mask Who Owns $1.4 Trillion of Treasuries. US, Argentina Sign $20 Billion Swap Deal Before Midterm Vote. How China Took Over the World’s Rare-Earths Industry. Kering Backs Away From Beauty With $4.7 Billion L’Oreal Deal. Runaway Insurance Costs Bring Back Talk of Price Caps. ECB faces lawsuit from staff union over ‘intimidation’ allegations. “If someone wanted to short the AI bubble, what would they sell?” Money, Women and Taxes: Jeffrey Epstein’s Fiery Friendship with a Wall Street Titan. How a Handyman’s Wife Helped an Hermès Heir Discover He’d Lost $15 Billion. The Fraudster Behind Steve Ballmer’s NBA Nightmare. “Inevitably, the most difficult family member will choose ‘family harmony’ as their most important value.” Brazen Heist at Louvre Museum Snatches Priceless Royal Jewels. James Fishback’s ETFs Are Being Liquidated Against His Will.

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!