What Are the Risks of RWAs? Tokenization does not create a new asset class — it builds a new stack of intermediaries around an old one, and the risks live at the seams.
"Real-world assets" bundles together a government bill, a rental house, a loan to a business in another country, gold in a vault, and a note referencing a share price. As investments those have almost nothing in common. What they share is a wrapper — and the useful way to think about the risks is therefore not asset by asset but layer by layer, asking at each level whether a risk was already there, whether tokenization changed its shape, or whether it exists only because a token exists. This report is that framework. For the underlying structures, start with what real-world assets are.
At a Glance Six layers: Asset, wrapper, parties, tech, exit, law Most risk is: Inherited from the asset Genuinely new: Keys, oracles, composability Tech failures: Mostly keys, not contract bugs Compensation schemes: Do not cover this Predictor: The less liquid, the worse 01 — Risk does not live in the token The single most useful correction is that almost nothing that goes wrong in this market goes wrong in the blockchain. Failures happen to borrowers, buildings, custodians, registries and the people holding keys. The token records the consequence faithfully.
A framework helps, and one word applied to each risk does most of the work. Ask whether it is inherited — it would exist if you owned the asset conventionally; reshaped — it exists in ordinary finance but behaves differently here; or introduced — it exists only because of the token. That classification is also, roughly, what the international financial-stability and securities bodies have concluded: their published assessments map tokenization onto existing risk taxonomies rather than inventing new ones, on the principle that the same activity should carry the same regulatory outcome. Most of this is old risk in new plumbing. One rule is worth stating in advance because the record bears it out: risk scales inversely with how liquid and observably priced the underlying asset is. And for readers outside the United States, one theme runs throughout — where you live changes what you actually own.
02 — The asset, and the wrapper around it The first layer is purely inherited. If the borrower fails, the tenant leaves, or the metal price falls, the token falls with it. Tokenization cannot make a weak credit strong or an empty building full, and no amount of settlement innovation touches the layer that determines whether you get paid. One sub-risk deserves naming: where an asset has no observable market price, somebody chooses the number — and the token inherits that choice, including into any system that later accepts it as collateral.
The second layer is introduced, and it is the most under-priced part of the whole market. There are three broadly different structures, and they are not equivalent. The issuer may put its own security directly on-chain; a third party may hold the real asset and issue a receipt against it; or a token may simply reference a price without any claim on the thing. In the second and third cases your claim is against the wrapper company, not the underlying issuer — a second, separate credit exposure that most buyers never price. Alongside it sits the rule that governs this entire market: in almost every jurisdiction, a ledger entry is not legal title. The land registry, the securities depository or the transfer agent's book remains authoritative, and the chain is evidence at best.
03 — The intermediaries you were told had gone The third layer is reshaped, and it inverts the original pitch. Tokenization was sold as disintermediation. In practice it adds intermediaries: a custodian, a transfer agent, a fund administrator, a tokenization provider, a market maker, and frequently an off-chain manager or originator besides. Several of these sit outside the regulatory perimeter that covers their conventional equivalents.
Two specific exposures follow. The first is concentration in small firms: tokenization providers are often young companies with short records and thin teams, and rating agencies have flagged exactly this as a distinct concern — a dependency on a handful of people that a conventional fund does not carry. The second is the third-party veto. An off-chain registry, a lender, or a counterparty can invalidate or block what the chain says: registries have refused to recognise tokenized claims, and a mortgage lender declining consent has been enough to kill a tokenized property transaction outright. The chain records intent; other people still control outcomes.
The blockchain is the most reliable component in the stack, and it is not where anyone has lost money. That has happened at the borrower, the custodian, the registry, and above all at whoever held the keys.
04 — The technology, and where it actually breaks The fourth layer is introduced — but not in the place most readers expect, and this is the single most useful empirical finding in the category. Logic bugs in smart contracts are a minority of losses. The dominant technical failure mode is compromise of keys and privileged access: the key that signs prices into an oracle, the key that can upgrade a contract, the credentials of an operator. Recent significant incidents in this sector occurred at protocols that had been audited more than once, because the audits were not the relevant control.
The conclusion worth carrying: an audit covers the contract — not key management, not oracle infrastructure, not custody, and not the fund. Treating "audited" as a safety claim mistakes a narrow technical review for an assessment of the business. A second exposure sits alongside it: whoever holds upgrade or pause rights can change the product's terms after you buy, which is a governance risk wearing the costume of code. Ask who those parties are before deciding the technology is neutral.
05 — Liquidity, in both directions The fifth layer is reshaped, and it fails in two opposite ways. The first is the mismatch: a promise of instant, continuous redemption written against assets that settle on business days at best, and not at all at worst. Under stress the wrapper cannot deliver liquidity the underlying does not have, and redemption windows narrow or close — a pattern sharpest where the underlying is genuinely slow to sell, as in real estate tokenization.
The second failure is more common and much less discussed: nothing trades at all. Dividing an asset into small pieces does not create anyone who wants to buy them. Several tokenized property platforms and security-token venues closed not because of fraud but because a secondary market never materialised, and where trading depends on a single market maker, it ends when that firm withdraws — the ordinary depth problem described in market depth. A third exposure completes the picture: exit terms can be changed. At least one product replaced an assumed at-par redemption with a lower floor, and the token repriced accordingly.
06 — Whose law, and whose courts The sixth layer is introduced by cross-border distribution, and it matters most to readers outside the country the product was designed for. Three points.
Classification differs by regime, so a licence is not a blanket. Some jurisdictions apply a facts-and-circumstances test; others draw a bright line, with the consequence that a venue's advertised authorisation is frequently the wrong licence for the specific product you are buying. Eligibility attaches to you, not to the token, and the token cannot tell where it is — the same instrument can be lawful for retail buyers in one place, restricted to professionals in another, and an unregistered offering if it reaches someone in a third. Acquiring it through a geographic workaround puts you outside the exemption's protections entirely. And recourse sits where the issuer is: disputes go to the courts named in the terms, often small offshore jurisdictions, frequently through arbitration, and enforcing a judgment against assets held elsewhere can cost more than the position is worth. On compensation schemes the honest answer is simply none — the investor-protection and deposit-guarantee arrangements that cover conventional brokerage and bank accounts do not extend here.
07 — What the record actually shows Four observations, stated directionally. First, the categories with the cleanest records are the most boring: through the middle of 2026 there is no publicly documented default, valuation loss or investor writedown at a major tokenized government-bill product or at the largest gold tokens — with the caveats that the history is short, the scale modest, the period benign, and that a derivative product built on top of a bill-backed token did fail badly. Absence of documented failure is not evidence of safety.
Second, the worst records belong to the illiquid, which is the rule from section 01 paying off: the credit and property categories account for nearly all documented losses, through borrower defaults, liquidity mismatches, and legal-layer failures. Third, the modal outcome is not catastrophe but irrelevance — a striking number of tokenization ventures, including bank-backed ones, simply closed for lack of volume. And fourth, the problems are structural rather than teething: a lender vetoing a tokenized property transfer in 2019 and tokens issued ahead of registry completion years later are the same failure, separated by most of a decade.
08 — What to ask, and of what These are questions to put to a document, not to a promoter. What am I actually holding — the asset, a receipt on a custodian, or a note referencing a price? Who is the issuing entity, where is it incorporated, and is it capitalised? Is the underlying segregated and bankruptcy-remote, with a named trustee and independent verification? Who can pause, freeze, upgrade or vary the terms — and have redemption terms ever been changed unilaterally?
Then: what is the exit — redemption at value, with whom, on which days — or only a secondary market, and who makes it? What feeds the price, and who controls those keys? What licence does the venue hold, and is it the right licence for this product's legal type? Am I eligible where I live, and does acquiring it another way void my protections? And what is the governing law and forum, and could I realistically litigate there? If the documentation does not answer these, that silence is itself the finding. The wider framing to keep is that supervision built around institutions is a poor fit for failures originating in infrastructure — which is precisely the shift tokenization makes, and precisely why the questions above are about parties and keys rather than about blockchains.
"He that diggeth a pit shall fall into it; and whoso breaketh an hedge, a serpent shall bite him."
— ECCLESIASTES 10:8
Methodology & Sources This report was prepared with a multi-agent research process covering the documented incident record across asset categories, the regulatory and legal position, and the risk taxonomies published by international financial-stability and securities bodies; claims that could not be substantiated were removed rather than hedged. It names no platform, protocol, issuer, custodian, provider or individual, and describes no incident in identifying detail. Figures are given directionally. The classification of risks as inherited, reshaped or introduced is an analytical device, not an official taxonomy, though it reflects the prevailing regulatory framing that tokenization largely maps onto existing risk categories under a same-activity, same-risk principle. The statement that no publicly documented default, valuation loss or investor writedown has occurred at major tokenized government-bill or leading gold products reflects the absence of publicly reported incidents as at the middle of 2026, is expressly not presented as evidence of safety, and is qualified by the sector's short history and benign operating period. Characterisations of failures elsewhere — borrower defaults, liquidity mismatches, registry and lender vetoes, venue closures for lack of volume, and compromises of keys or privileged access at audited protocols — reflect publicly reported events; where facts are disputed or allegations unadjudicated they have been excluded rather than repeated, and no conclusion is drawn as to fault in any matter. Regulatory descriptions are general and jurisdiction-dependent; specific classification, licensing and eligibility outcomes vary and continue to evolve, and readers should verify against current official sources. The observation that investor-compensation and deposit-guarantee schemes do not extend to these instruments is general and readers should confirm their own position. Nothing here is a recommendation regarding any product, issuer or transaction, and nothing here is legal, tax or investment advice.