I’ve spent years telling people that Coldcard and other hardware wallets sitting in a drawer, disconnected from the internet, are about as safe as crypto custody gets.
I still believe that’s mostly true. But the disclosure this week forced me to update that advice with an uncomfortable footnote: even a device explicitly built to never touch the internet can still generate a broken key, silently, for years, without anyone noticing until real money started disappearing.
What Actually Went Wrong Inside Coldcard’s Firmware
Coinkite, Coldcard’s manufacturer, confirmed the underlying problem directly, disclosing a critical entropy-generation flaw affecting Coldcard firmware, according to Coinkite’s own advisory.
The vulnerability reduced the effective randomness behind seeds generated on certain devices, and the company was blunt about the fix: update to patched firmware, generate an entirely new seed, and migrate funds away from anything touched by the old one, since a patch fixes future key generation but does nothing to repair a seed already created with weak randomness.
Independent security researcher 0xQuit broke down the technical root cause shortly after the exploit began, explaining that bad randomness in Coldcard’s seed generation process let roughly 594 BTC get swept from around 500 addresses in a short window.
He noted the device’s PIN, air gap, and secure element didn’t help once the seed itself was compromised, and recommended anyone holding significant value move to multi-device multisig using at least two different vendors, writing seeds only on paper and splitting them geographically using Shamir secret sharing.
Block’s Investigation Found The Attack Was Still Live
The most detailed technical breakdown came from Block, the parent company behind the competing Bitkey wallet, whose leadership confirmed their engineering and security teams began investigating the moment reports surfaced, according to Block’s official disclosure. Block was upfront that no Block products were affected, but continued investigating specifically to protect the broader self-custody community. What they found was serious: the Mk2 and Mk3 firmware intended to use a hardware random number generator, but a mistake in a firmware macro caused it to rely instead on a known device ID, timer state, and call history, making wallet generation deterministic rather than genuinely random. Newer models attempted to compensate at boot, but a reseed process truncated that input down to just 32 bits, still far below what a wallet should have.
Bitkey’s engineering lead added a genuinely useful detail during the same investigation, noting an unusual pattern in the sweeps led the team to confirm the attacker used a paid account at a well-known blockchain-services provider to query source addresses during the theft, according to a follow-up technical disclosure. I think that’s a meaningful operational security mistake on the attacker’s part, the kind of detail that gives investigators a genuine thread to pull.
CZ’s Blunt Warning: Nothing Is 100%
Binance founder Changpeng Zhao weighed in directly once the scale of the theft became clear, writing plainly that even hardware wallets and long-established wallets can have bugs, and asking crypto holders to consider splitting their funds across a few wallets as mitigation, while acknowledging that approach carries its own risks, according to his public statement. He closed with his familiar refrain: stay informed, stay SAFU, and remember that nothing is 100% safe.
I think CZ’s honesty about the tradeoff is worth sitting with. Splitting funds across multiple wallets and vendors genuinely reduces concentration risk, but it also means more seed phrases to track, more devices to maintain, and more opportunities to lose track of something during recovery. Diversification isn’t a free upgrade, it’s a different risk profile entirely.
Alternative One: Multi-Vendor Multisig, Done Correctly
The first alternative worth taking seriously is multisig, splitting signing authority across multiple independent hardware devices from different manufacturers, so a flaw in any single vendor’s product can’t unilaterally move your funds. 0xQuit’s own recommendation was specific: generate at least two seeds using at least two different vendors, configure a multisig using wallets from each seed, and keep those seed backups physically separated.
I don’t think multisig should be treated as an automatic fix, though. It only protects you if every signer independently verifies the actual transaction details on a trusted display, rather than trusting a single interface to show them the truth, since a compromised signing interface can still trick multiple honest signers into approving something malicious without any single device itself being broken.
Alternative Two: MPC Wallets That Never Create A Single Seed Phrase
The second alternative worth serious consideration, especially for larger holdings, is multi-party computation, or MPC. Rather than generating one private key that gets split or backed up after the fact, MPC distributes the actual signing process across multiple independent parties, so no single device, person, or location ever holds a complete key at any point. There’s no single seed phrase anywhere in the system, which means there’s no equivalent of Coldcard’s entropy bug sitting in one component waiting to be exploited.
I think MPC’s real advantage here is structural. The failure mode that hit Coldcard, a single flawed randomness source silently undermining every key built on top of it, simply can’t happen the same way when key generation itself is distributed across independent parties from the start.
Alternative Three: Layering Your Own Independent Entropy
The third approach doesn’t require abandoning hardware wallets entirely. It means adding your own independent entropy source on top of whatever your device generates internally, rather than trusting a single manufacturer’s random number generator completely, whether that’s rolling dice, drawing cards, or combining several separate entropy sources and mixing that manually with your hardware wallet’s own output.
I think this appeals to people who don’t want the added complexity of multisig or MPC but still want a genuine hedge against exactly the kind of silent firmware bug that just hit Coldcard. It takes more diligence and it’s easy to get wrong if you’re not careful about how you actually combine the entropy sources, but done properly, your security no longer depends entirely on one company’s chip and code being flawless for years on end.
The Real Lesson Isn’t Which Product You Choose
I don’t think the takeaway here is that Coldcard specifically is untrustworthy going forward, or that any single brand is inherently safer than another. CZ’s advice after this incident was simple: split funds across multiple wallets, devices, and vendors, so a flaw in any one of them only ever exposes part of your holdings rather than everything at once. Copying the same seed phrase across several devices doesn’t achieve this, since a flaw in that one seed compromises every device holding it simultaneously. True diversification means genuinely separate seeds, generated independently, ideally on hardware from different manufacturers entirely.
Multisig, MPC, and manual entropy layering all address the same underlying problem from different angles: never let a single point of failure, whether that’s one company’s firmware, one seed phrase, or one device, be the only thing standing between your assets and a bug nobody catches until it’s too late.
Disclosure: This is not trading or investment advice. Always do your research before buying any cryptocurrency or investing in any services. Follow us on X @nulltxnews
Be the first to comment