Fifteen attackers, and the number is climbing. Over the past 72 hours, the Bitcoin blockchain has become a silent battlefield. According to Galaxy Research, more than $130 million in BTC has been drained from Coldcard hardware wallets – wallets that were once described as the 'gold standard' for self-custody. The attackers are not exploiting a complex DeFi hack or a compromised smart contract. They are simply brute-forcing private keys generated with insufficient entropy. And the window for remaining victims is closing with every block.
I remember when Coldcard became the go-to recommendation for serious Bitcoiners. It was the answer to every question about security. No screens? No problem. It is air-gapped. It is open-source. It is the choice of the paranoid. And now we know that paranoia was not enough. The hardware that was supposed to be the last line of defense has become the attack surface itself.
Coinkite, the Canadian manufacturer, has confirmed that its firmware for the Mk2, Mk3, and Mk4 models used MicroPython's software Pseudo-Random Number Generator (PRNG) in certain seed-generation routines, instead of the hardware true random number generator. The result: entropy levels of around 40–72 bits, far below the 128-bit minimum we expect from a secure wallet. That is not a subtle error. That is a catastrophic failure of the most fundamental security assumption in cryptocurrency: the generation of a private key that cannot be guessed.
Let me explain what this means in practice. A 128-bit key, if truly random, has more possible combinations than there are atoms in the visible universe, approximately. A 40-bit key has about 1.1 trillion possibilities. That sounds like a lot until you realize that a modern GPU can try billions of combinations per second. We are not talking about a sophisticated nation-state attack. We are talking about any reasonably equipped hobbyist. The Bitcoin blockchain is a public ledger. Every address that has ever received funds is visible. If your address was generated by a Coldcard with low entropy, an attacker can scan the entire UTXO set, identify weak addresses, and derive the private key using basic computational resources. The first wave of thefts occurred just hours before Coinkite's public announcement.
This is not a remote exploit or a supply-chain attack. This is a consequence of a firmware design that silently fell back to a software PRNG when generating seeds. The hardware security module that should have provided true randomness was bypassed, either because of a broken integration or a lazy fallback in the boot path. Coinkite has pushed a hotfix to all affected models and release channels, but the hotfix cannot repair seeds already generated. The only remedy is to move your bitcoin to a new wallet. Every affected user must perform an emergency migration, or risk losing everything.
When I first read the Galaxy Research report, my immediate reaction was not surprise but resignation. During my years auditing protocol architecture – starting with that 2017 deep dive into the 0x relayer design – I learned that the most dangerous vulnerabilities are rarely the exotic ones. They are the mundane choices that nobody questions. A random number generator is one of those mundane choices. We spend so much time worrying about elliptic curve math and side-channel attacks that we forget to ask: where do the random numbers come from? The answer, in this case, was a piece of interpreted code running on top of bare metal.
I have consulted for institutional funds that pride themselves on rigorous security. In 2024, when I helped a UK pension fund draft its Bitcoin investment thesis, we spent hours debating the philosophical merits of self-custody versus ETF custody. The fund eventually allocated 2% to the asset, but only after an extensive review of key management. We never asked whether the hardware wallet's entropy source was certified. We assumed that a reputable vendor with a history of security evangelism would get the fundamentals right. This incident proves that assumptions, even reasonable ones, are the enemy of security.
The on-chain evidence is still unfolding. Galaxy Research has reportedly received reports from 73 victims, but the actual number is likely far higher. Many long-term holders have not yet checked their wallets. Some may not even know they are at risk. The report estimates that more than 7,300 wallets could be affected, containing over $130 million. The attacker count is growing daily, because anyone with the technical ability and a few hundred dollars in electricity can join the hunt. The barrier to entry is zero, and the reward is extraordinary.
Let me be precise about the technical details. The affected Coldcard models – Mk2, Mk3, and Mk4 – all suffered from this entropy issue. The vulnerability lies in the seed generation routine, which apparently calls MicroPython's os.urandom or an equivalent software PRNG when the hardware RNG is not properly initialized. This is not a deliberate backdoor; it is a bug that should have been caught in basic code review. But the consequence is identical to a backdoor. The private key space is so small that any attacker can enumerate all possibilities and check for matching public keys on the blockchain.
There is a deeper issue here that the initial reporting has missed. The fallback to a software PRNG suggests that the hardware security element's true random number generator was not correctly integrated into the firmware's boot path. This is not something you fix with a one-line patch. It requires a re-architecture of the secure element interface, a revalidation of the entire seed generation process, and a thorough audit of every code path that touches randomness. Coinkite will need to do far more than issue a hotfix; it will need to rebuild trust from the ground up.
The market response, so far, has been muted. Bitcoin's price has not moved significantly, which is unsurprising given that $130 million represents a small fraction of daily trading volume. But the legacy effects are more subtle. The 90% of stolen funds that have not moved represent a potential supply overhang. Attackers are not dumping; they are holding. That suggests a degree of sophistication – they are either waiting for market liquidity, planning a laundering operation through mixers or cross-chain bridges, or simply treating this as a long-term investment. The on-chain addresses are known, and we should be monitoring them closely. The protocol remembers what the market forgets.
For the self-custody ecosystem, this is an existential moment. Hardware wallets have been marketed as the ultimate expression of 'not your keys, not your coins'. But the Coldcard incident reveals a corruption in that slogan. Your keys are only as secure as the entropy that created them. If the entropy is predictable, your keys might as well be written on a sticky note. The industry's response to this should not be a whisper campaign against Coldcard. It should be a collective reevaluation of how we verify the security claims of every hardware vendor.
I have spent the last decade advocating for self-custody. I have argued that exchanges are dangerous intermediaries and that holding your own keys is a moral imperative. Yet events like this make me hesitate. The emotional toll on victims is not just financial; it is a betrayal of a core belief. I know that feeling intimately. In 2022, after the Terra collapse and the Celsius bankruptcy, I retreated to a cabin in the Scottish Highlands for six weeks. I watched the industry that I had championed betray its own promises. The burden of belief became almost unbearable. That experience taught me that solidarity with victims is not about empty platitudes. It is about demanding systemic accountability.
The contrarian angle that many will resist is this: the Coldcard attack is not a reason to abandon self-custody. It is a reason to demand a new standard of proof. We should not run back to custodial exchanges, which have their own track record of incompetence and theft. Instead, we should push for a 'Proof of Entropy' – a verifiable attestation from any hardware wallet that the seed generation process used a certified true random number generator, with no fallback path. This is not a rhetorical demand. It is a technical specification that can be implemented and audited.
We can also learn from the attackers' behavior. They are not indiscriminately emptying wallets. They are selectively targeting addresses with high balances, and they are holding what they take. That indicates a strategic mind, perhaps a group that understands Bitcoin deeply. We should not underestimate them. The fact that 90% of the stolen funds remain untouched suggests they are not in a hurry. They are waiting for the right moment to move, and when they do, we will see it on-chain. We must have monitoring infrastructure in place to flag their wallets and alert the community.
As a protocol product manager, I have learned that resilience is not a property of a single node. It is a property of a network. The same principle applies to self-custody. Relying on one hardware wallet brand, one firmware version, one entropy source – that is not resilience, that is centralization of trust. The solution is diversity: generate seeds on multiple independent devices, verify the output with a separately sourced random generator, and store the resulting mnemonic in geographically distributed locations. This is more work, but the cost of laziness has just been measured at $130 million.
Let me also address the regulatory dimension. Law enforcement agencies around the world are investigating this crime. Victims are being urged to report their losses to local and federal officials. That is important, but we should not expect a complete recovery. The more likely outcome is that some attackers are caught, some funds are frozen, and a stark reminder is sent to the industry. Regulators may begin to ask whether hardware wallets should be subject to security standards, similar to how financial institutions are regulated. That could be a positive development, if it forces vendors to perform independent audits and disclose vulnerabilities in a timely manner.
Coinkite's response has been mixed. The co-founder, Rodolfo Novak, issued a public apology, which is commendable. The hotfix was pushed quickly, which is also good. But the company has not yet published a detailed technical post-mortem. That is a mistake. In the absence of transparency, the community will assume the worst. I have seen this pattern before: a security incident is announced, a patch is released, and then everyone moves on to the next drama. But the victims never move on. They are left with the task of migrating funds and praying that the attackers have not already found them.
The most urgent advice for any Coldcard owner is simple: stop using the affected device immediately. Do not merely update the firmware, because updating does not repair the seed. Create a new wallet on a trusted device, transfer your bitcoin to it, and then discontinue the old wallet. This is a painful process, especially for those who have followed best practices like using passphrases and multi-sig. But it is the only way to eliminate the risk. If you are reading this and are not sure whether your Coldcard is affected, assume it is. Check your generation date, check the firmware version, and if in doubt, move your funds.
For the rest of the ecosystem, this event should be a call to action. We need to fund open-source tools that detect weak entropy in real time. We need to encourage disclosure of entropy source details in every hardware wallet specification. We need to create a certification body that tests these claims with independent hardware analysis. This is not over-engineering; it is basic fidelity to the promise of self-custody. Trust is not given; it is verified. And verification must be continuous, not a one-time seal of approval.
I will end with a reflection that has been forming in my mind since the news broke. The Coldcard incident is not a story about a single company's failure. It is a story about the danger of iconography. We elevate certain brands to the status of talismans, believing that their name alone protects us. But code is not magic. It is written by humans, and humans make mistakes. The only defense is a culture of rigorous, open, and continuous audit. We build in silence so the network can speak. But silence, too, can be a lie when derived from a flawed entropy source.
The protocol remembers what the market forgets. The blockchain has already recorded the addresses that were drained, and that record will exist forever. It will be a permanent lesson for anyone who might forget the importance of randomness. In the years ahead, we will look back at this event and see it as a turning point – not because Bitcoin failed, but because we finally acknowledged that security is not a product, but a practice. The attackers are still active. The funds are still at risk. And the window is closing. The question is not whether Bitcoin survives this challenge. The question is whether we have the humility to learn from it, the courage to change our habits, and the wisdom to demand more from the tools we trust with our wealth.
Freedom arrives when the gatekeepers go dark. But freedom also requires that we understand the gates themselves. Hardware wallets are the gate-keepers of our individual sovereignty. When they fail, the failure is not theirs alone. It is a collective failure of our imagination. We imagined that we could purchase safety. We cannot. We can only build it, together, day by day, line by line of code, and bit by bit of entropy.
The Coldcard drama is not over. As of this writing, fifteen attackers have drained over a hundred million dollars, and more are joining every hour. The victims are someone's parents, someone's children, someone's savings. The best we can do now is to stop the bleeding, help those affected, and ensure that the next generation of hardware wallets is held to a standard that would make an attack like this impossible. That is the work ahead. And it is, without exaggeration, the most important work in self-custody today.
