For educational purposes only; not investment advice. Investing may result in loss.
Direct answer
On-chain credit attempts to establish lending credit using address history, collateral, identity or real assets. This entry explains the difficulties of DeFi unsecured loans.
On-chain Credit is a mechanism that uses blockchain records, verifiable identity, collateral relationships and repayment history to evaluate a borrower’s ability to perform. It attempts to expand DeFi from the over-collateralization model of “deposit $150 worth of assets and lend out $100” to more capital-efficient credit lending. However, public transaction records only indicate what the address has done, and do not automatically indicate who is behind the address, whether they are willing to repay in the future, and whether there is recourse after a breach of contract.
Credit is not a score, but a judgment of future solvency and willingness to pay. The on-chain system can observe address balances, loan periods, liquidation history, governance participation and interactive agreements; the off-chain system can also introduce corporate statements, income, legal identity, bank statements and guarantees. Different solutions simply combine these signals in different ways.
On-chain credit is roughly divided into four categories. The first category is still collateral-centric and only adjusts the collateral rate based on historical performance; the second category is address reputation, which grants quotas based on continued use and repayment behavior; the third category relies on real-name identities or non-transferable certificates to associate multiple addresses to real entities; the fourth category consists of capital pools providing loans to specific institutions, market makers or real enterprises, and recourse through legal contracts.
True unsecured loans are not “without risk control”, but that risk control shifts from on-chain assets that can be liquidated instantly to income, reputation, guarantees, legal enforcement and risk pricing. If these alternative constraints are unreliable, the protocol simply replaces transparent collateral risk with opaque credit risk.
How it works
Permissionless addresses can be created for free, and one person can control hundreds of addresses. If the protocol issued a $1,000 loan based solely on a new address, the borrower could take the funds and abandon the address, with the protocol neither knowing their identity nor recourse. This is the dilemma created by the combination of the Sybil problem and the lack of recourse.
Over-collateralization changes the judgment of “whether the person is trustworthy” to “whether the collateral is sufficient.” When the value of the collateral falls close to a threshold, the smart contract allows the liquidator to repay the debt and take away the discounted collateral. Rules can be automatically enforced and are suitable for permissionless environments, but are less capital efficient and may still generate bad debts in the event of price gaps, oracle anomalies, or chain congestion.
Credit loans need to answer five questions: who is borrowing the money, whether the data is true, how to determine the amount, whether the interest rate covers expected losses, and who will perform recourse after default. A simplified framework can be used to understand interest rates:
Loan interest rate ≈ cost of funds + expected default loss + operating and liquidity compensation
If the probability of default is 4%, the recovery rate after default is 25%, and the expected credit loss is approximately 4% × (1 - 25%) = 3%. That does not include administrative costs and uncertainty. The model results depend on the sample and the economic cycle and cannot be regarded as a fixed answer.
Example
Assume that an on-chain fund pool has 20 million USDC and provides one-year revolving quotas to three market-making institutions. Borrower A has a credit line of US$5 million with an annual interest rate of 11%. Borrower A will first provide US$500,000 of junior, loss-absorbing capital and a contract will be signed by a legal entity. The loans are not completely unsecured, but have lower collateral coverage and rely mainly on corporate assets and legal recourse.
The agreement estimates that A’s one-year default probability is 6%, the default recovery rate is 40%, and the expected loss is 3.6%, or $180,000. The nominal interest rate of 11% is approximately $550,000, after deducting management fees, idle funds, audit and potential recovery costs. If a sharp market decline causes multiple borrowers to default at the same time, the historical average probability will become invalid and the capital pool may freeze withdrawals for a long time.
Six months later, A’s on-chain balance dropped from 8 million to 1.2 million US dollars, but the exposure to the centralized exchange cannot be observed. The risk committee put its new drawdowns on hold, but existing loans of $4 million have not yet matured. Smart contracts cannot be liquidated immediately like collateral lending; fund providers can only wait for repayment, negotiate a restructuring, or go through legal proceedings. On-chain accounting improves transparency, but does not eliminate the real credit cycle.
Risks
Funding providers should examine borrower concentration, single industry exposure, collateral and security quality, past due history, legal jurisdiction, audit frequency and withdrawal queues. The APY displayed in the agreement is the result of borrowing interest distribution, not the risk-free interest rate. If 40% of the funds are lent to the same entity, the impact of a default is much greater than the scattered average data.
Also distinguish between “interest accrued” and “cash received”. Some pools will include interest receivable in the net worth even if the borrower has not paid; after an overdue payment occurs, the page income may lag behind the true loss. Bad debt provisions, write-off rules and management suspensions should be made public.
Common misconceptions
Myth 1: Transparent history on the chain equals reliable credit
Addresses can be changed, transactions can be brushed, and cross-chain and off-chain liabilities may also be invisible. Transparent records are just inputs and still require subject identification and data quality.
Myth 2: Unsecured loans have no collateral risk, so they are safer
It reduces collateral price fluctuations, but increases borrower default, information asymmetry, recovery cycle and legal enforcement risks. The risks are only transferred rather than disappeared.
Myth 3: High loan rates are enough to cover any bad debts
Defaults are correlated, and multiple borrowers may fail at the same time in a crisis. No matter how high the accrual interest rate is, it cannot make up for the large area of principal that cannot be recovered.
Myth 4: DAO approval represents professional risk control
Voters may lack credit experience or have conflicts of interest. Specific processes, accountable entities and ongoing monitoring should be assessed rather than just the form of governance.
Related topics
Sources
- DeFi Lending: Intermediation Without Information? - BIS (accessed: 2026-08-21)
- IRB Approach: Treatment of Expected Losses and Provisions - Basel Committee on Banking Supervision (accessed: 2026-08-21)
- Default Process - Goldfinch (accessed: 2026-08-21)