5.3 Counterparty and Credit Risk
Why tokenization leaves the borrower's credit risk in place and adds the token issuer's on top, what the First Brands and Tricolor collapses show about where credit losses start, who chooses borrowers in on-chain credit and who absorbs the loss, how few firms stand behind the role count, how DeFi collateral stacks counterparties, and where wrong-way risk sits.
Tokenization does not change who owes you money. It adds to the list of parties who must pay before you do. A tokenized loan pool is owed by the same borrowers as the paper one. Around that obligation the token adds an issuer or SPV, a platform that keeps the register, a custodian, and a cash-leg issuer, and once the token is posted as collateral in DeFi, a lender, a curator, and sometimes a second token issuer. The Basel Committee defines counterparty credit risk as "the risk that the counterparty to a transaction could default before the final settlement of the transaction's cash flows"1. Every party on that list is a counterparty in that sense, and none of them removes the borrower's risk.
Section 5.2 established that a token proves control, not ownership, and traced what a holder's legal claim is when a custodian or SPV fails. This section moves from whether the holder owns the claim to whether the parties behind it can pay, and keeps two promises from Chapter 4. Section 4.1 recounted the on-chain private credit losses and concluded that "Credit risk did not disappear on-chain; it moved to whoever chose the borrowers." This section builds the framework on those incidents without retelling them. Section 4.3 found that role bundling concentrates the dependency chain into fewer hands than the role count suggests and wrote "That concentration is Section 5.3's subject"; the fifth heading below takes it up.
Tokenization Changes the Wrapper, Not the Borrower
The institutions that have examined tokenized funds apply the existing credit analysis and add technology risk on top. None subtracts credit risk. Moody's wrote, in a report quoted by Crane Data, that tokenization "does not change the core credit risks of a money market fund or short-term bond fund," and that what it can add is "novel risks, and in particular operational risks arising from blockchain use"2. The BIS finds that tokenized money market funds "give rise to risks that mirror, and potentially amplify, those found in conventional money market funds"3.
The same holds further out on the credit curve, where Section 2.2 placed the harder assets. The FSB's May 2026 report on private credit describes the borrowers a tokenized credit pool is likely to hold: they "typically lack public ratings," where rated they "are often rated around single B-," and "Certain lending practices may obscure true leverage"4. A token over such a book carries that borrower. Figure's HELOCs are recorded on Provenance, yet still go into securitization trusts in the ordinary way, with Figure "required to retain a certain portion of the securitization trusts into which the loans are sold" in order "to satisfy risk retention regulations"5. The ledger records the lien; the credit work is what any HELOC securitization gets.
Tokenization can remove one counterparty risk, and it sits in settlement. Section 3.4 showed that atomic delivery versus payment eliminates principal risk between trade and settlement. That gain concerns whether both legs of a trade arrive together, not whether the obligor behind the asset leg pays.
The Losses Start Off-Chain
The most instructive credit failures of the past year were not on a blockchain. First Brands, an auto parts maker, and Tricolor, a subprime auto lender and ABS issuer, both collapsed in September 2025. Prosecutors and the SEC allege collateral that did not exist or had been pledged more than once. According to the Southern District of New York, by about August 2025 "Tricolor had pledged approximately $2.2 billion of collateral to lenders and investors, but Tricolor had only approximately $1.4 billion of real collateral"6. The SEC's complaint, announced on 18 August 2026, alleges that Tricolor represented the loans in its ABS pools as "free and clear of any other liens when the defendants knew that many had been or would soon be double pledged," and that "more than $945 million of principal" from those offerings was outstanding at the bankruptcy7. Two First Brands executives are charged with schemes to "double- and triple-pledge loan collateral" and to sell factoring partners "billions of dollars of purported customer receivables that did not exist"8. These are allegations, and the Justice Department's releases state that the defendants are presumed innocent unless and until proven guilty68.
The official readings point at opacity. The FSB found that "While fraud appears to have played some role, a lack of transparency and hidden leverage characterised these failed companies"4. The ECB drew the same lesson: the defaults "have illustrated how weak underwriting standards and opacity can transmit losses across parts of the financial industry"9.
No tokenized product is known to have held First Brands or Tricolor exposure. The point is structural: a token over a pool of auto loans, receivables, or ABS inherits the pool's blind spot. If the originator has pledged a loan twice, the token records a claim that someone else also holds, and the token's contract cannot see the other pledge. Tokenizing a claim is not verifying it.
A ledger can help under one condition. IOSCO records the suggestion that DLT-based repos "could improve the degree of visibility regulators have in relation to the sources and uses of securities in collateral agreements"10, and Figure's DART registry already records liens on Provenance (Section 4.1). A second pledge becomes visible only if the on-chain registry is the authoritative lien record and every lender checks it. A token that mirrors an off-chain pool sees what the originator reports.
Two Obligors Where There Used to Be One
A token adds the credit of whoever issues it. Section 5.2 showed that every layer between holder and asset adds a balance sheet, and quoted SEC staff's warning that third-party token holders may be exposed to the third party's bankruptcy. The FSB puts the same point in credit terms: a token's price may reflect "the credit risk of a token issuer" as well as the reference asset11. IOSCO is sharpest on the synthetic token, where "counterparty risk exposure is to the token issuer to fulfil its obligations instead of the underlying asset issuer, which might not always be clear to investors"10.
Sponsor support is not an obligation. Franklin's prospectus for its on-chain money fund says: "The Fund's sponsor is not required to reimburse the Fund for losses, and you should not expect that the sponsor will provide financial support to the Fund at any time, including during periods of market stress"12. It is standard money fund disclosure and the right baseline. RealT, whose title structure Section 5.2 examined, shows the sponsor side: the sponsor managed every property, and when the City of Detroit sued, a court, not the token, ordered tenants' rent into escrow; the city's claims remain allegations. Where loss protection exists, it is contractual and sized. USDY launched with a first-loss position of about 3% (Section 4.3); Tinlake pools had a junior tranche below the senior DROP tokens, and Maple's pool cover was mostly depleted by the late-2022 defaults (Section 4.1). The question for a holder is what absorbs the first dollar of loss, how large it is, and who funds it.
Each added party also adds time. When a borrower behind an SPV defaults, the holder's recovery runs through the borrower's workout first and the SPV's distribution second; Harbor Trade's senior position sat in a court-supervised workout for more than a year (Section 4.1). Whether the claim survives an issuer's insolvency as property is the question Section 5.2 answered. This section counts how many balance sheets must stay solvent for the holder to be paid.
The Underwriter Is Paid; the Holder Takes the Loss
On-chain credit delegates the choice of borrower to a third party. Section 4.1 found that every loss in its history came from lending "underwritten by a delegated third party, with the protocol standing outside the eventual workout." As a framework, that pattern becomes three questions for any tokenized credit product: who chose the borrowers, how that party is paid, and what it loses if it chose badly.
The FSB framed the risk in securitization terms: token issuers "could become analogous to originators of securitised assets in the traditional context and could introduce hidden (i.e. not disclosed or properly priced) risks for investors," particularly where securitisation rules, "including credit risk retention or disclosure rules," do not apply to them11. Risk retention exists because an originator that sells all of a loan has less reason to care whether it is repaid. Figure, which keeps a slice of its trusts under those rules, shows the old discipline carrying over. DeFi's newer intermediaries may not.
Morpho's documentation defines a curator as the party that "configures vault risk parameters and liquidity allocation rules, abstracting risk curation decisions away from depositors," and states that "Curators set performance and management fees together with the corresponding recipients"13. The page describes how curators are paid and says nothing about curators bearing losses. Depositors hand over the risk decision and keep the risk.
FalconX's Credit Vault on Pareto names M11 Credit as curator and says it "performs credit evaluations, enforces loan covenants, and manages real-time risk monitoring." The vault lends to "OspreyX 2024-A Limited, a bankruptcy-remote SPV designed to separate investor capital from FalconX's corporate balance sheet," and "FalconX provides a first-loss capital contribution"14. The borrower keeps some of the loss, though the page does not say how much. That answers the third question better than a fee-only curator does, but each holder still has to read for it, product by product.
Ratings do not settle the question. The FSB observed that private credit ratings are "sometimes provided by smaller lesser-known agencies," and of First Brands and Tricolor that "Lender reliance on overly optimistic credit ratings may have also contributed to a flawed assessment of underlying risks"4. A rating says what one agency concluded, not who chose the loans.
Fewer Hands Than the Role Count Suggests
Read as a credit question, bundling means that one firm's failure removes several links in the chain at once, and that clients who never deal with one another become linked through it. The FSB warned that the failure of a large tokenization platform "could have negative spillover effects across a range of financial institutions"11. IOSCO adds that concentration risks are "significantly exacerbated where investors are exposed to singular entities for the conduct of activities across the entire value chain"10.
In its July 2026 registration statement Securitize calls itself "the only vertically integrated tokenization provider with SEC-registered entities across a transfer agent, broker-dealer, alternative trading system (ATS), investor advisor and fund administration"15. It was the largest distributed platform on rwa.xyz on 28 September 2026, at $4.68 billion of $38.61 billion, about 12% by this paper's calculation16, in line with Section 4.5. The dependence runs both ways: "We may overly rely on certain products (such as BUIDL, which represented more than 60% of Securitize's tokenized assets under management as of September 30, 2025)"15.
The filing also shows how far bundling can reach. From November 2024 through 2025, Securitize lent to investors in tokenized funds against their tokens, locked that collateral in its sToken vaults so it could "interact with various decentralized finance protocols," and borrowed on-chain against it, acting "in its capacity as a regulated intermediary and transfer agent for the underlying collateral assets"15. The registrar was also lender, collateral holder, and the party re-posting the collateral into DeFi. The program was small, $1.7 million of restricted tokenized assets at 31 December 2025, and was wound down by 31 March 202615. It illustrates structure, not a live exposure: legally separable roles sat in one entity, over the same tokens.
Banks concentrate across both legs. BNY is custodian of BUIDL and USTB and administrator of BUIDL (Section 4.3). Circle announced in March 2022 that BNY Mellon "was selected as a primary custodian for USD Coin (USDC) reserves"17, and BNY was selected in July 2025 for primary custody of RLUSD reserves (Section 4.4). No bank here is shown to be weak; the point is that a holder spread across products and cash tokens may be less diversified at the custodian than the count suggests.
Figure sits at the far end: originator, servicer, and securitization sponsor for its HELOCs, the firm that built the ledger they are recorded on, and keeper of the DART lien registry (Section 4.1). Its filing notes that adverse findings about its practices "could lead to downgrades by one or more rating agencies of the bonds issued by our securitization trusts"5. A problem at the servicer reaches the bonds. On 28 September 2026 the Figure HELOC token was $23.19 billion of the $36.51 billion of represented credit on rwa.xyz, about 64% by this paper's calculation18.
No integrated tokenization platform has failed, and bundling saves the reconciliation between register and chain that Section 4.3 described. The FSB and IOSCO warnings describe a failure the market has not yet produced.
Collateral on Collateral: How DeFi Stacks Counterparties
Once an RWA token is posted as collateral, every hop adds a counterparty, and looping adds the same obligor again. Section 3.4 argued that composability transmits stress in general; this part names the credit chains.
The FSB set out the mechanism: "Rehypothecation of tokenised assets could lead to the build-up of leverage in the financial system," and "there may not be adequate regulation in place stipulating minimum haircuts"11.
The chains already exist for Treasury funds. IOSCO found that BUIDL's largest holders in June 2025, "USDtb ($1.3 bn); Spark ($800 mn); and Ondo ($150 mn)," "were firms which used BUIDL tokens as collateral assets in their products and services"10. The BIS notes that allow lists "only constrain direct holding" of tokenized money market funds, and that "the practice of repackaging TMMF shares into other tokens can also transmit stress from one fund to others"3. For Treasury funds the underlying credit is strong and the layers matter mostly for liquidity. For credit funds, the layers stack credit on credit.
FalconX's Credit Vault shows the credit version. Depositors fund the vault in USDC, the vault lends to FalconX's SPV, and FalconX's lending desk "uses the vault capital to extend overcollateralized credit to institutional clients such as hedge funds and trading firms"14. The vault token then returns to DeFi. In Gauntlet's levered strategy, "The vault deposits CV tokens as collateral on Morpho and borrows USDC against it. The borrowed USDC is used to buy additional CV tokens, which are again deposited as collateral on Morpho"19. Morpho states that at launch the market "supported loan-to-value up to 77%"20; at that ratio, repeated loops can take a position toward about 4.3 times the looper's own capital, by this paper's calculation. The vault reached a record $144 million on 23 June 2026, and FalconX, citing Token Terminal, called its token the second-largest RWA collateral on Morpho, at nearly $76 million on 22 June 202614.
The diagram reads top to bottom as the path of credit exposure from a DeFi lender to the borrowers who ultimately repay. Green marks the parties supplying money, navy the structures it passes through, and red the obligors everything above depends on. It simplifies: the looper can be a curated vault or an individual, and many depositors never loop.
The pattern extends to other credit funds. Morpho's playbook describes a loop for Apollo's sACRED, already quoted in Section 4.4. Stablecoin ecosystems also fund credit directly: Anemoy announced in June 2025 that Grove, "the institutional credit platform incubated in the Sky ecosystem, is seeding the fund with a massive $1 billion allocation," the fund being JAAA, Janus Henderson's tokenized AAA CLO fund21. JAAA held $587.1 million on 28 September 202618, so the launch figure is not a current holding. The FSB's private credit report already finds leverage "at multiple and varying levels" and warns that "This layering effect may amplify losses during market stress"4. DeFi looping adds one more level on top.
Wrong-Way Risk: When the Collateral and the Counterparty Fail Together
The Basel Committee defines specific wrong-way risk as exposure to a counterparty that "is positively correlated with the probability of default of the counterparty due to the nature of the transactions with the counterparty," and general wrong-way risk as default probability "positively correlated with general market risk factors"1. IOSCO puts the standard in collateral terms: ideal collateral would "not correlate with the provider's credit risk"10. No tokenized RWA loss from wrong-way risk has been documented yet, so the argument is from structure. Three patterns stand out.
The first is a single-name credit token used as collateral. FalconX names the event every layer of the chain above is exposed to: "it is possible for volatile collateral to drop faster than liquidations are possible, potentially leading to impairment of the vault. This could also impact any secondary markets for the vault token"14. A sharp crypto sell-off stresses FalconX's clients, its loan book, the vault token, and the looper's position at once, so for the Morpho lender the counterparty's default risk and its collateral's value move together. That is general wrong-way risk, and looping multiplies it. How fast the token's price would show an impairment is Section 5.1's question.
The second is a platform holding its largest client's product as its own cash. Securitize reported that "18.0 % of their cash and cash equivalents are BUIDL as of December 31, 2025"15, the product behind more than 60% of its tokenized assets under management three months earlier. Securitize adds that it "believes it is not exposed to any significant credit risk on cash"15, and BUIDL holds Treasury bills and cash, so the credit exposure is small. The correlation is elsewhere: the filing notes that BlackRock can terminate its placement agent agreement "at any time with prior written notice"15, and an event that disrupted BUIDL or that relationship would reach the platform's revenue and its cash together.
The third is stablecoins and tokenized funds that back each other. Section 4.4 showed tokenized money market funds becoming stablecoin reserves, as with USDtb, while OUSG's instant redemptions depend on Circle's USDC. The BIS warns of "the potential for shocks to propagate between the TMMF and stablecoin markets"3. The credit on both sides is mostly Treasury bills; the channel is liquidity.
Each pattern raises a question existing rulebooks were not written for. Do risk retention and disclosure rules reach a token issuer that selects loans, as the FSB asked? Do haircut rules reach a DeFi market that accepts a single-name credit token at 77% loan-to-value? Who supervises an integrated platform as a whole rather than licence by licence, and who answers for a curator's choice of borrower? Chapter 6 takes up how regulators are approaching them.
Credit risk decides whether the money is there. Liquidity risk decides whether a holder can get it out, and in what order, when others want out too; a credit event is often what starts the queue. Section 5.4 takes up liquidity risk in secondary markets: the gap between redemption and a two-sided market, the mismatch between a liquid token and an illiquid underlying, and what happens to both under stress.
- Tokenization changes the wrapper, not the borrower. Moody's states that it "does not change the core credit risks" of the funds it assessed, and the BIS finds tokenized money market funds carry risks that "mirror, and potentially amplify" those of conventional funds.
- The losses start off-chain. The FSB found "a lack of transparency and hidden leverage" behind the 2025 First Brands and Tricolor collapses, with collateral allegedly pledged two or three times. No tokenized product is known to have held either, but a token over such a pool inherits the blind spot.
- A token adds an obligor rather than replacing one. IOSCO, the FSB, and SEC staff each warn that holders can carry the token issuer's credit or bankruptcy risk on top of the underlying's, and Franklin's prospectus says its sponsor "is not required to reimburse the Fund for losses."
- Fewer firms stand behind tokenized products than the role count suggests. Securitize, by its own filing, has SEC-registered entities across five roles, had more than 60% of its tokenized assets under management in BUIDL in September 2025, and for a time lent against client tokens and re-posted them into DeFi. The FSB warns that a large platform's failure could spill across the institutions using it.
- DeFi collateral stacks counterparties and can build in wrong-way risk. A credit-vault token backed by lending through one prime broker was looped on Morpho at up to 77% loan-to-value, so the lender's collateral and its counterparty weaken in the same sell-off.