In the first two articles, we discussed the importance of private keys, as well as the trade-offs between self-custody and custodial wallets. But if you were expecting me to tell you, “If you choose A, you’ll be safe,” or “If you choose B, you’ll be foolproof,” I may have to disappoint you—because in the crypto world, there is no absolute safe solution, only risk management. We’ll use two real cases to illustrate this.
FTX collapse: a textbook-level risk of centralized custody
In November 2022, the world’s second-largest cryptocurrency exchange, FTX, collapsed overnight. Investigations showed that Alameda Research siphoned roughly $8 billion from FTX customers’ funds for high-risk investments and to plug shortfalls. Customers deposited their coins into FTX, believing that under a “custody” model their assets would be safe—only to find those coins were treated as the exchange’s own funds and used accordingly.
The lesson from FTX reveals a fundamental problem: when you hand asset custody to an institution, you essentially place your trust in the institution’s management, risk-control systems, and audit transparency. If all these safeguards fail—misappropriation by insiders, related-party transactions, financial fraud—no matter how advanced the technology is, it can’t protect you.
Afterward, the industry reached a consensus: exchanges must provide proof of reserves, regularly publish on-chain wallet addresses, and allow users to verify whether the platform’s assets can cover users’ deposits. Industry leaders like Binance were among the first to implement this mechanism. It’s not a perfect solution, but it significantly raises the bar for wrongdoing.
Coldcard was stolen: self-custody isn’t a silver bullet
You might say, “Alright, I’ll handle everything myself—then it’s definitely safer, right?” The Coldcard case may make you rethink that.
Coldcard is considered one of the safest hardware wallets on the market, known for “military-grade security.” However, at the end of 2024, a user bought a brand-new Coldcard wallet through official channels. After receiving it, they initialized it as usual and transferred a Bitcoin amount. A few days later, the assets were stolen.
The findings are shocking: the attacker didn’t attack the wallet itself—they attacked the supply chain. The attacker infiltrated a distribution link and, after the device left the factory but before it reached the user, embedded malicious firmware. The user thought they were using a “cold wallet,” but the attacker had already intercepted the private key at the moment it was generated.
The Coldcard case shows that self-custody security is far more than just “keeping the seed phrase safe.” The attack surface also includes:
Whether the purchase channel is tampered with (supply chain attacks)
Whether the firmware signature is verified (malware disguised as official updates)
Whether the environment that generates the seed phrase is safe (networked devices may be monitored)
Whether the receiving address is altered during transfer (clipboard-hijacking attacks)
The attack surface shifts, not disappears
By comparing the FTX and Coldcard cases, you’ll notice a core pattern: when you put your assets in an exchange, the attacker targets the platform; when you manage them yourself, the attacker targets you. The attack target shifts from “the platform” to “the individual,” and the attack surface shifts from “institutional internal risks” to “the supply chain, operational processes, and device security”—the risk doesn’t disappear, it just moves.
The probability of a black swan event at the scale of FTX is extremely low, but if it happens, it’s catastrophic. The risk of self-custody, however, is everywhere, all the time—every transfer, every device connection, every copy-paste could be an opportunity for an attack.
Conclusion: the essence of security is risk management
There’s no “one-click security” button in the crypto world. Real security means understanding where the risks you face actually lie—and then making trade-offs.
If you’re a typical user, you accept the centralized risks of custodial services and choose a platform with audit transparency and a proven track record. If you’re a hardcore user and insist on self-custody, then you need to handle the entire stack properly: hardware procurement, firmware verification, multi-signature setup, and secure physical backups.
There is no absolute secure solution—only proactive choices made after fully understanding the risks.
Next up: Cold wallets, MPC, multi-signatures—how exactly do these custodial technologies protect assets?