The Implications of CeFi and DeFi in Bankruptcy: A Hot Take on Celsius

The difference between owning (ie having property rights in something) and being owed (ie having a right against a person by way of obligation) is at its most stark in bankruptcy. In the wake of multiple bankruptcies of crypto entities last year, the question of whether customers of these entities owned or were merely owed has come into sharp focus, especially when the entity has been engaged in what the industry has coined CeFi (centralised finance), which involves the lending and borrowing of cryptoassets. In theory, the same questions are engaged in a DeFi (decentralised finance) transaction, where one of the roles is substituted by a peer (often pseudonymous) rather than a centralised intermediary. Determining whether property in ‘loaned’ cryptoassets belongs to a failed borrower (whether centralised intermediary or peer) who simply owes an obligation to the lender or remains with the lender throughout involves a question of characterisation. Two recent motions in the Celsius litigation are instructive of how this characterisation is undertaken by the courts and although the case involves CeFi rather than DeFi, there is no reason why the same reasoning would not extend to DeFi as well.

The first motion pertained to customers with assets in Celsius’ ‘Earn’ accounts, who had been promised as high as 18% interest on sums deposited. Despite their strenuous arguments to the contrary, the bankruptcy court decided that those assets had become the property of Celsius, giving effect to Terms of Use stating that those customers had ‘grant[ed] Celsius … all right and title to such Eligible Digital Assets, including ownership rights’. This outcome in truth was always inevitable, perhaps even in the absence of such clear language.

The arrangement mirrors exactly the traditional position relating to interest-bearing bank accounts, established since the mid-19th century in the common law world to entail an obligation owing to customers rather than ownership of money by customers. Cases like Foley v Hill (1848) 2 HLC 28, 9 ER 1002 and Thompson v Riggs 72 U.S. 663 (1866) decided that absent any special conditions, money deposited with a bank became the property of the bank, with the customer obtaining a claim to the return of the same amount of money (plus interest) on demand. The obligation is fungible (the bank need not return the same coins and notes deposited) and forms the basis for borrowing short and lending long that is the backbone of modern banking. This characterisation is necessary because money, unlike some other property, does not have any inherent utility except by way of alienation. A bank uses deposited money as it pleases (usually by extending loans itself), makes what profit it can, and pays back to its customer the principal and agreed interest. How could a bank generate profits off the money were it otherwise?

It is the same with most cryptoassets such as Bitcoin. Whether described as staking, liquidity mining, or onward lending, all these processes involve committing the assets in a manner inconsistent with the continued ownership of the original owner of the assets. In the first motion, Celsius’s customers argued that the Terms of Use stated that they had ‘loaned’ the Earn assets to Celsius, claiming this meant that they retained ownership of those assets. Terminology has never been conclusive in an exercise in legal characterisation, but this argument was doomed to fail for a more basic reason. The term ‘loan’ has two different meanings. The customers argued that they believed they had made a ‘loan’ in the non-fungible sense of the term (the exact thing must be returned), as where one lends a chattel whose possession itself has utility. Many examples abound. A book can be read and enjoyed; a car can be used to ferry passengers for a profit. Such ‘loans’ can be, and are often structured, as not involving the transfer of ownership from lender to borrower. No library transfers ownership of library books to patrons in return for a mere obligation for their return. Likewise, hire agreements of cars are careful to spell out that ownership of said vehicles remains with the hire company throughout.

However, most cryptoassets are not suitable for non-fungible loans since, like money, they can only be enjoyed through alienation. As Warren Buffett famously explained in a CNBC interview, ‘If you said … for a 1% interest in all the farmland in the United States, pay our group $25 billion, I’ll write you a check this afternoon,’ Buffett said. ‘[For] $25 billion I now own 1% of the farmland. [If] you offer me 1% of all the apartment houses in the country and you want another $25 billion, I’ll write you a check, it’s very simple. Now if you told me you own all of the bitcoin in the world and you offered it to me for $25 I wouldn’t take it because what would I do with it? I’d have to sell it back to you one way or another. It isn’t going to do anything. The apartments are going to produce rent and the farms are going to produce food.’ Without treading on the controversial question of whether all or most cryptoassets are thus Ponzi schemes, the point holds that simply holding cryptoassets doesn’t generate any returns so, unless a loan of cryptoassets is characterised in the same way as a loan of money, it is difficult to see what the point of the loan is (from the borrower’s point of view) and how the borrower is supposed to generate the profits in order to pay the lender the interest it has promised. As a result, it is unsurprising that the court found that the customers had clearly made a ‘loan’ of the fungible rather than non-fungible variety, where ownership of the property passed to Celsius in return for an obligation to return an equivalent sum plus interest.

More fortunate than the Earn customers were a group of customers who had placed cryptoassets in Celsius’ ‘Custody’ accounts. On 20 December 2022, Judge Glenn granted a motion authorising their withdrawal of, among others, certain digital assets which had ‘only ever’ been held in those accounts. The crucial difference was that the Terms of Use stated that ownership in cryptoassets deposited in the ‘Custody Program’ would ‘at all times remain with the [user]’ and that Celsius would not ‘transfer, sell, loan or otherwise rehypothecate’ such assets. The bankruptcy court agreed that assets in such ‘Custody Wallets’ did not form part of the Celsius estate. This again is undoubtedly correct. Though the legal principles relating to ownership of cryptoassets remain unsettled, in the absence of countervailing commercial indications and outside of the security context, the passing of title should be governed by the intention of the parties, such consensualism being consistent both with longstanding authority (see eg Cochrane v Moore (1890) 25 QBD 57; Metropolitan Trust Co of New York v McKinnon 172 F. 846 (1909)) and contemporary theories of justice in transfer.

One theme that emerges from these two motions is the importance of the characterisation process: intention is the starting point but the commercial nature of the transaction cannot be ignored. Yet, characterisation is distinct from questions of actual segregation, which may pose independent obstacles to recovery. In the Celsius case it seems that customer assets were in fact segregated in the ‘Custody Wallets’. The ‘Custody’ claimants were remarkably fortunate in this respect. Although ownership is superior to obligation in bankruptcy, its advantage is contingent on the customer’s ability to identify the property owned. Some crypto entities, such as FTX, are alleged to have been treated customer assets as belonging to them even though their terms of use indicate otherwise, in which case they would likely have been long dissipated or irretrievably commingled. In such cases, even if custody is taken to mean the retention of ownership, some aspects of crypto-systems may thwart customers’ claims. Given the pseudonymity and immutability of blockchains, establishing the ownership of property you are unable to identify or recover will likely prove to be cold comfort.

Kelvin Low is a Professor at the National University of Singapore.

Timothy Chan is a Sheridan Fellow at the National University of Singapore.

This post first appeared in the Harvard Law School Bankruptcy Roundtable Crypto-Bankruptcy Series on 14 March 2023.