The Wrong Question About Crypto Platform Risk

Most days in crypto start the same way. You open X or Reddit and there’s another hack, another platform in trouble, and another fight over what it means for the industry. A few hot takes later, you have a new narrative, a few headlines, and the same debate we’ve had a hundred times before.

I’ve worked with financial and crypto products for more than a decade. I’ve seen TradFi, CeFi, DeFi, and CeDeFi. My career journey is something for another article, but during that decade, I’ve seen the industry move between two fairly predictable conclusions. Centralized finance is too opaque, so DeFi fixes it. Then DeFi gets hacked, so DeFi is too risky, and centralized platforms are safer. But then a centralized platform collapses.

We’ve been seeing this for years, and I think the problem is partly the question itself. Even the platform where you read my thoughts on has a special “CeFi vs. DeFi” tag. “CeFi or DeFi?” tells you about how a platform works, but surprisingly little about how many things have to work correctly for your money to remain safe. And that second question becomes very interesting as financial platforms get bigger.

CeFi Learned the Complexity Lesson the Expensive Way

Celsius and BlockFi are good places to start because neither failed simply because it was centralized.

Crypto came in from customers. Once assets leave a client’s own wallet, they become part of the platform’s financial ecosystem. From there, what happens to them depends on how that platform operates: assets could then be lent, pledged, reinvested, or exposed to other counterparties. Those counterparties had their own counterparties, collateral requirements, liquidity needs, and market exposure. Add leverage and maturity mismatches on top. Now, a product that looks, if not simple, at least fairly familiar from the user’s side reveals an even more overwhelming financial structure underneath. 

Basically, that’s what rehypothecation does to the risk map. I gave the platform my one asset, and it participates in several economic relationships at once. When one part fails, everyone suddenly becomes very interested in exactly where the collateral went. Especially me, since it was my collateral in the first place. BlockFi’s problems, for example, became tied to exposure to Alameda Research and FTX. If BlockFi’s stability is no longer dependent only on BlockFi, its users are dealing with problems from several relationships away. It’s a fairly complex topic, and we took a deeper look at it together with CryptoSlate’s analysts in a dedicated guide on collateral reuse risk.

And this isn’t exotic by any means. Traditional finance has spent a very long time discovering creative ways for counterparty relationships to become someone else’s problem. Crypto just managed to run through some of the same lessons at much higher speed.

There are also less dramatic ways for centralized platforms to disappear. In July 2026, BitMEX announced that its exchange would close in September after what the company described as a strategic review of the business and the broader crypto industry. BitMartannounced an orderly wind-down around the same time, including the phased discontinuation of spot trading, futures, staking, lending, and other products it put money into.

BitMart may be the case everyone is looking at today, but the pattern matters more than the name. There are now unproven concerns around WOO X’s liquidity as well. I wouldn’t treat that as a fact until there is stronger evidence. If another exchange runs into the same problem, we’re looking at a pattern. Maybe WOO X turns out to be fine. At least I hope it does, but I also wouldn’t ignore the signal completely.

So, users think about risk in terms of one question: Can this platform be hacked or become insolvent? Businesses have a much longer list: Can it remain compliant and maintain banking and liquidity and operate in every market it serves? Can its counterparties and the company itself make enough money to keep doing all of that? And the bigger the business and more interconnected the system becomes, the more ways there are for something to go wrong.

DeFi Removed the Company. Did It Remove the Risk?

The company in a CeFi model is often described as an ultimate point of failure. You don’t have to trust in an intermediary. That’s compelling, because instead of asking a platform what happened to the money, I can just inspect transactions on-chain. Self-custody also removes the need to hand permanent control of assets to an intermediary. Permissionless protocols operate without an employee deciding whether you qualify to use them.

These are real advantages that remain a staple in crypto. I don’t think it makes sense to dismiss them just because DeFi has security problems. But removing the intermediary doesn’t remove the system around the transaction. In CeFi you build a castle with a Castle Lego kit. In DeFi everyone shares what they have and goes with the flow.

So. In DeFi I don’t depend on a company, its employees, and its counterparties. I depend on smart contracts, price oracles, bridges, governance systems, external protocols, validator infrastructure, private keys, and the economic assumptions connecting all of them. Ouch.

2026 has a lot of examples of that kind. Kelp DAO lost roughly $292 million in Aprilafter attackers compromised infrastructure around its LayerZero-powered bridge. One of the more interesting details is that the smart contracts themselves apparently did what they were supposed to do. The false information entered elsewhere in the system, and perfectly functioning contracts acted on it.

Earlier that month, Drift Protocol lost roughly $285 million. Again, attackers reportedly spent months building relationships with contributors and eventually obtained approvals that gave them administrative control. Once they had it, the protocol behaved accordingly.

In August, Tectonic was hit for an estimated $75 million after an attacker manipulated the price of an illiquid token and used the inflated asset as collateral. Cronos eventually halted the blockchain to prevent more funds from leaving.

Humanity Protocol had a different problem in June. The attacker compromised a director’s device, obtained operational keys, and used that access to execute unauthorized actions across Ethereum and BNB Smart Chain.

Then there are governance attacks. Term Finance lost an estimated $8.5 million in August through its vault governance system. It had a seven-day timelock and a mechanism allowing liquidity providers to veto proposals. The protections existed but still didn’t prevent the attack.

Moonwell’s August incident involved another combination: oracle-price manipulation and collateral accounting, with security firms estimating losses at around $8.7 million.

I don’t know about you, but we are not even through 2026 yet, and these incidents alone account for more than $700 million in reported losses.

Transparency Is Not the Same as Simplicity

With that much of a loss, a “DeFi hack” is a convenient phrase, but it hides how different these incidents actually are. 

1. Smart contract exploits attack flaws in the protocol’s code or logic.
2. Oracle manipulation attacks the information a protocol uses to decide what an asset is worth. Lending protocols are particularly sensitive to this because collateral can be perfectly real while its reported price is completely wrong.
3. Bridge exploits attack the infrastructure connecting assets or messages across chains. Kelp DAO is an unusually good example because the downstream consequences reached protocols whose own contracts had not been compromised.
4. Private-key and infrastructure compromises don’t necessarily break the protocol at all. If an attacker obtains legitimate administrative authority, the blockchain may simply execute legitimate-looking instructions from the wrong person.
5. Governance attacks exploit the mechanism intended to give a decentralized system collective control. If voting power, permissions, or proposal mechanics can be captured, governance itself becomes an attack surface.

And then there is composability, which connects all of the above. Aave didn’t need its core contracts to be directly hacked to feel the Kelp DAO incident. The attacker was able to use rsETH inside Aave, turning a problem that originated elsewhere into a liquidity and bad-debt problem for another protocol.

A transparent complex system is still a complex system. Knowing that a protocol depends on five smart contracts, two external price feeds, a bridge, another lending market, and token governance is better than not knowing, for sure. But is it simple? I doubt it. Composability is simultaneously one of its best features and one of the reasons its risk can be difficult to map. There is no one to blame and no penny to find.

Audits have the same limitation. They are extremely useful, but an audit tells you something about the code and assumptions examined at a particular point in time. It cannot guarantee that another protocol won’t fail, an oracle won’t report bad data, governance won’t make a bad decision, an admin key won’t be compromised, or a new integration won’t create an interaction nobody anticipated. I mean, every decision makes sense individually, but the system gets wonky. I’ve seen a similar mistake outside crypto many times in consulting. A company adds another product because customers want it, another integration because growth needs it, another partner because the economics are better, and another process because the previous three now need coordination.

The Scale Paradox

Large platforms offer deeper liquidity, more products, more markets, and more integrations. Large DeFi protocols can build enormous pools of capital and become infrastructure for dozens of other applications. Large centralized platforms can negotiate better counterparties, maintain larger compliance teams, and spread operational costs across millions of customers.

Scale creates so many dependencies. In CeFi, those dependencies include lenders, custodians, liquidity providers, market makers, banking partners, payment providers, and other financial institutions. In DeFi, we are talking about contracts, bridges, oracle networks, governance systems, validators, liquidity pools, and external protocols.

This creates a slightly uncomfortable scale paradox: the infrastructure can become more capable at the same time that the number of things capable of breaking increases. I don’t think the conclusion is that large platforms are inherently unsafe. It’s that smaller infrastructure with fewer dependencies can be easier to reason about and reveal fewer points of failure. The conclusion is that complexity itself belongs in the risk calculation. Users already do this instinctively with financial products. If two products produce roughly the same result, but one requires three intermediaries and the other requires twelve, you probably want to know why the other nine are necessary. Crypto platforms deserve the same question.

Can a Smaller Platform Keep the Risk Map Smaller Too?

When I first started working at CoinRabbit in 2023, one thing took some adjustment after years of consulting for much larger and very different businesses: there simply weren’t that many layers. But working inside a smaller one has made me think more, not less, about what centralization actually needs to involve. A smaller platform cannot realistically build every product, support every possible financial strategy, and become infrastructure for half the industry. But I’m increasingly convinced that’s not necessarily a bad thing. We build new products and features, but we don’t try to solve every problem at once. We focus on what our clients actually need and develop from there.

CoinRabbit is centralized. Maybe this is where my perspective isn’t so neutral since I work here. But I do know that clients trust us with assets, even though that creates counterparty risk by definition. Our approach has been to keep the financial structure relatively narrow: fewer counterparties, fewer dependencies, and no rehypothecation of client collateral. Instead of trying to make deposited collateral productive somewhere else in the system, we keep it reserved.

This is also where Austrian economic theory starts to shape how I think about the product. Money doesn’t become more productive just because you create more claims on it. If a client expects their assets to be available while those same assets are deployed somewhere else earning another return, you now have two economic expectations around the same capital. 

Repeat that across enough layers, and the whole structure starts depending on one assumption: everyone won’t ask for their money back at the same time. That can work perfectly well until it doesn’t. And when one layer runs into liquidity problems, the problem can travel through the others surprisingly fast. Traditional banks at least have central banks and other liquidity mechanisms behind them when this happens. Crypto doesn’t have an equivalent safety net. So adding another layer of capital reuse isn’t free efficiency. You’re adding another dependency, and I think that dependency should be included in the risk calculation. 

We prefer a more predictable setup. If a client gives us collateral, we protect it. If a client explicitly wants to put their assets to work and earn interest, that’s a different decision, which is what our savings product is designed for. There is a business reason for this approach too. We want clients to stay with us for years, so it makes more sense to build relationships that last than to maximize what we can earn from any single position. Some of those relationships started long before I joined CoinRabbit, and I hope they continue for many years to come.

At that level, the conversation is rarely just about rates. Clients want to know what else we can do for them around the core product. Can you structure a cross-collateralized loan position? Can you put some extra protection around my loans while I’m away on vacation with my family? Can you help me move some funds into fiat if I don’t have a local bank account yet? Can you help me finance a house purchase using my Bitcoin? These questions are often more important than the headline rate.

That last question sounds almost embarrassingly simple after several years of crypto trying to automate everything. In DeFi, automation is the product. A smart contract doesn’t know that you’ve been a client since 2020, that you’re moving assets between wallets, or that your situation today is unusual. The rules apply without discretion, and that’s great. It also means there may be nobody on the other side who can look at the situation and help. A smaller CeFi platform introduces trust in a company, but it can also give clients fewer operational layers and an actual person who understands the account.

Maybe CeFi vs. DeFi Isn’t the Most Useful Question

After enough years in crypto, I’m suspicious of any argument that ends with one architecture being “the safe one.” Neither model removes risk. They put risk in different places, though. Celsius and BlockFi illustrated what can happen when centralized financial relationships become opaque and interconnected. DeFi demonstrated that replacing those relationships with code doesn’t magically eliminate dependencies: a) smart contracts can fail, b) oracles can be manipulated, and c) bridges can break or governance can be captured. And sometimes every individual component works exactly as designed while the system as a whole still produces a very expensive result.

So maybe the more useful question isn’t whether a platform is centralized or decentralized.

Maybe it’s: How many things have to go right for my assets to remain safe?

We are talking: Who controls the assets? Where are they held? What can happen to them after deposit? How many counterparties touch them? Which external systems does the platform depend on? What happens when one fails? And, if something genuinely unusual happens, is there anyone capable of intervening? Those questions work for CeFi. They work for DeFi. They also work regardless of how large the logo on the homepage is.

For years, crypto platforms competed on more, more, more. I suspect the next stage will involve learning how to do less. Especially fewer things that can go wrong.

Source: finbold.com