Why this matters
Self-custody is supposed to be the safest way to hold Bitcoin. The Coldcard exploit is a reminder that "not your keys, not your coins" only holds if the keys themselves were generated correctly — and for five years, some of them weren't.
Confirmed losses tied to the exploit have grown from an initial $38 million estimate reported in late July to more than $130 million across roughly 7,300 addresses, according to Galaxy Research. At least 15 independent attackers are now believed to be exploiting the same flaw, turning what began as a handful of coordinated thefts into an open scramble to drain whatever wallets remain unmigrated.
What went wrong
The root cause traces back to a March 2021 firmware release for the Coldcard hardware wallet, made by Canada-based Coinkite. That firmware generated recovery seeds with far less randomness than intended — roughly 40 bits of effective entropy instead of the 128-bit standard a secure seed requires.
A gap that small is not a rounding error. It means an attacker who understands the device's timing behavior can enumerate the possible seeds offline, without ever touching the physical wallet, and eventually reconstruct the private keys attached to it.
Affected models include certain Mk3 devices running firmware 4.0.1 or later, along with older firmware on Mk4, Mk5, and Q units. Wallets generated using Coldcard's dice-roll option, which sources randomness independently of the firmware's internal generator, are not affected by the flaw.
Four waves and counting
The theft has unfolded in stages rather than as a single event. A first wave drained tens of millions of dollars before the flaw became public knowledge. Later waves picked off wallets whose owners hadn't yet heard the news or hadn't acted on it. A fourth wave, still active this week, swept several hundred additional BTC from addresses that had gone unmigrated.
Each wave has attracted more attackers rather than fewer. Once the vulnerability was documented publicly, it stopped being the property of whoever found it first. Independent researchers, opportunistic hackers, and at least a few automated scripts are now all racing to claim funds from the same finite pool of exposed wallets — which is part of why the loss estimate keeps climbing days after the initial disclosure.
What Coinkite has done, and what it hasn't fixed
Coinkite has shipped emergency firmware for every affected model and says it has destroyed its remaining vulnerable inventory before it could reach new customers. That closes the door for future buyers.
It does nothing, however, for a seed that was already generated on the flawed firmware. Updating the firmware on a device whose seed is already compromised does not change the seed. Anyone in that position is not looking at a vulnerability to patch — they are looking at a private key that may already be known to someone else, whether or not it has been swept yet.
A timeline of an escalating story
- Late July — First reports surface of a Coldcard firmware flaw, with early loss estimates around $38 million.
- August 1–2 — Estimates climb past $89 million as researchers identify additional affected wallets and a second wave of thefts begins.
- August 3 — A fourth wave of draining activity is documented, with losses reported near $116 million and hundreds of newly compromised addresses.
- August 4 — Galaxy Research puts confirmed losses at 1,596 BTC (over $100 million) across roughly 7,300 addresses, with a suspected total near 2,055 BTC (about $130 million) once the unconfirmed fourth wave is included.
Each new estimate has represented additional wallets being found and drained, not just a restatement of the same theft. That pattern — a slow-motion disclosure rather than a single breach — is part of what has made the exploit so hard to contain. Publicity that is meant to warn users also tips off new attackers to a vulnerability they hadn't previously known how to exploit.
Hardware wallets aren't magic
It's worth being precise about what this exploit does and doesn't say about hardware wallets as a category. The flaw lived in Coldcard's specific seed-generation code, not in the general concept of storing keys offline on a dedicated device. Other hardware wallet makers have had their own security incidents over the years — supply-chain tampering, phishing campaigns impersonating support staff, and firmware bugs of varying severity — but a five-year-old entropy failure of this scale, affecting seed generation itself rather than a downstream feature, is unusual even by that track record.
The broader lesson is that a hardware wallet's security rests on code most owners never read and could not audit even if they tried. Buying a well-reviewed device is necessary but not sufficient. Verifying firmware versions, understanding how a given wallet generates randomness, and knowing what independent verification options exist (like Coldcard's own dice-roll seed option) are the parts of self-custody that get skipped once the device is out of the box and working.
What affected users need to do
Coldcard's own guidance, echoed by security researchers tracking the exploit, is unambiguous: a single-signature wallet whose seed was generated on vulnerable firmware cannot be made safe by an update alone. The seed itself is the compromised asset. Funds need to move to a wallet built from an entirely new seed — generated on patched firmware, ideally using the dice-roll method — or to a reputable custodian while a new setup is arranged.
Multi-signature setups where only one key came from an affected device carry lower risk, since an attacker also needs the other keys to move funds. Lower risk is not the same as no risk, and multisig users with an exposed key should still rotate it.
Bitcoin Gate Take
This is the ugly side of self-custody: the tooling can fail in ways ordinary users have no way to audit themselves, and five years is a long time for a flaw like this to sit undiscovered. We are not walking back our view that holding your own keys beats trusting an exchange with them. But self-custody has to mean verified, entropy-checked custody, not just owning a device with a reputable name on the box. If you set up a hardware wallet years ago and haven't checked how its seed was generated since, this week is the week to check — not to assume someone else already flagged it for you.
Bitcoin Gate's course covers seed generation and self-custody security in detail, if this story left you wanting to double-check a wallet you haven't opened in a while.