Skip to content
Legal

Risk summary

A short, complete list of what can cost you money here — including the risks for which no mitigation exists at all.

This page is a summary. The full risk section of this documentation covers each item in depth with worked scenarios. If you read only one legal page, read Nature of the collateral; if you read two, read this one as well.

The four that have no mitigation

Risk

1. Issuer or custodian failure. If the collateral issuer or its custodian fails, the price feed keeps publishing the price of the underlying share while the token's value goes to zero. The protocol cannot detect this and would keep lending against it. Nothing on-chain covers it.


Rule R-15.4.1 (8)

Risk

2. A freeze or blocklist on the collateral token. If the issuer can freeze, blocklist or upgrade the token's balances, the protocol's collateral could be immobilised. The automatic response to a contract upgrade — minting frozen, liquidations suspended, a transferability self-test, shutdown at 28 days — does not cover a freeze of the protocol's own balance. Nothing does.


Rules R-2.4.4, R-4.3.9

Risk

3. An unanticipated corporate action. A spin-off can resume the price materially lower with the spun-off entity never credited on-chain. Positions become liquidatable at the new price after the return ramp. A definitive loss of 5–30% is possible, and no human safeguard exists because none can be added.


Rules R-11, R-15.4.1 (9)

Risk

4. An atomic exploit. A bug exploitable in a single transaction is not covered by anything. The Closer key responds at human latency; against an atomic attack, human latency is irrelevant. What defends against this class is design — no flash mint, a one-block rule between risk increases, no external call in the mint or repay path — not a key and not a pause.


Rule R-12.1.8

Protocol risks with partial mitigation

RiskWhat limits it
Liquidation at any hourConservative out-of-session valuation, two-source confirmation of a fall, and the guarantee that repaying and depositing collateral are never blocked. Not a guarantee against liquidation.
Bad debt on a branchThe Stability Pool absorbs first, then direct liquidators, then a Backstop capped at 2% of debt, then pro-rata redistribution across that branch only. Branches are isolated; the loss stops at the branch.
A ceiling that is too generousThree independent caps, each read on-chain: measured market depth, a share of on-chain market capitalisation, and a dated absolute step that locks permanently the first time bad debt is ever recorded.
Concentration$50,000 per address for the first 90 days, then a formula. Circumventable by using many addresses; it makes concentration visible, not impossible.
Correlated collateralSPY, QQQ, VOO and IVV are close to the same asset. A shared hourly liquidation budget and a global ceiling limit throughput. Diversification is not available on day one.
Oracle failureGraded degradation, never substitution. Stale for 7 days shuts the branch down. There is no fallback oracle by design; a wrong price is worse than no price.
USDG depegThe PSM cap of 30% of supply, which never rises, and a temporary entry freeze while the Closer key exists. After day 365, nothing reacts.
Redemption of your positionChoose a higher rate, delegate, or repay. You are redeemed at oracle price and the fee stays with you; you lose exposure, not value.
Sequencer outage or censorshipDetected only if an uptime feed exists on this chain, which is an open verification item. If none exists, the risk is documented and not mitigated.

Rules R-6.2.1, R-6.8.4, R-5.4.1, R-5.4.4, R-5.4.3, R-6.3.3, R-6.9.1, R-8.6.1, R-15.4.1 (4)

Immutability, in both directions

Guarantee

Nobody can raise your rate, seize your collateral outside the published rules, block your repayment, blacklist your address, redirect the interest, or change a parameter. The functions do not exist in the bytecode.


Rules R-12.2.1, R-12.2.3

Risk

And nobody can fix anything. A wrong parameter is wrong forever. A failing component closes its branch instead of being replaced. A documented instance of what this costs: a protocol with no governance was drained of roughly $775,000 over two days in August 2026 and, in the operator's own words, "because the contract is immutable and has no governance, there is no pause button: users have to withdraw themselves."

Risk

Article 1(5b) of MiFIR is the largest unresolved question in this project. The collateral is a transferable security. A regulation in force since 2024 requires that all multilateral systems in which third-party buying and selling interests in financial instruments can interact operate as a regulated market, an MTF or an OTF. Whether a CDP's liquidation engine, redemption mechanism and Stability Pool constitute such a system has never been analysed in any published work. A dedicated memorandum from a financial-markets firm is commissioned before mainnet, and the answer is not known today.

Risk

MiCA. fyUSD referenced to one dollar would be an e-money token if it had an issuer. The position taken is that an immutable contract with no upgrade key is not a natural person, a legal person or an undertaking, and therefore that fyUSD has no identifiable issuer within the meaning of the regulation — which places it outside Titles II, III and IV under recital 22. This position has never been tested before a competent authority. The precedent for what happens when a stablecoin issuer is identifiable and established in the EU is unambiguous and it is not favourable.

Risk

Personal exposure of contributors. A US court has held it adequately alleged that a governance-participating collective operates as an unincorporated general partnership with joint and several personal liability. "No entity" is not protection; it is the absence of protection. This is a risk to the people building the protocol, and indirectly to anyone relying on their continued existence.

Phase risk

The contracts are new. Total borrowing is capped at $2M in the first phase, per-address positions at $50,000 for 90 days, and each branch's ceiling opens in dated steps that lock permanently if bad debt is ever recorded. Two branches are live at launch; three are dormant and activate only when on-chain metrics clear fixed thresholds.

Those caps are the honest expression of the residual risk in new code. They are not a marketing device, and no key can lift them.

Rules R-5.4.3, R-5.4.4, R-5.4.7, R-12.3.2

Total loss

You can lose everything you deposit. That sentence is not softened anywhere in this documentation and no page contradicts it.

Last reviewed: 2026-09-07 · Spec v0.4