Imagine waking up one day, checking your cold wallet, and discovering that your saved Bitcoins have disappeared without you having pressed a single button. This is exactly what has happened to thousands of Coldcard users, one of the most prestigious hardware wallet brands in the crypto world.
At the end of July 2026, a software error caused by old firmware triggered a massive drain that is already shaping up to be the largest security incident in the history of physical wallets, with over $111 million stolen.
Faced with a blow of this magnitude, the first doubt that arises in the community is inevitable: was the Bitcoin network compromised? The short and forceful answer is no. What is truly important about this case is understanding that it was not an attack on Bitcoin, but a flaw in the firmware of a specific manufacturer that had been hidden for more than five years.
The Bitcoin protocol functioned with absolute normality at all times, validating technically correct transactions. The problem resided exclusively in the local tool, whose randomness when creating keys was predictable.
Coldcard and the Role of a Hardware Wallet in Bitcoin Custody
To understand what happened, the first thing is to be clear about the difference between the protocol and the tool:
- Bitcoin:It is the global and decentralized network. Its rules are validated by thousands of independent nodes around the world.
- Coldcard:It is a physical device created by the Canadian firm Coinkite. Its role is to generate and store your seed phrase offline. It does not modify the network; it only safeguards your keys.
This cold storage strategy seeks to isolate your data from viruses and cyberattacks, making Coldcard the preferred choice for the most stringent investors.
However, its security depends on something vital: randomness. A weak random number generator is like a lock that always falls on the same combinations. If the initial seed is created with a predictable pattern, the physical isolation of the device is useless.
That is why the impact was so strong: the flaw shook confidence in self-custody and prompted many to hastily move their funds, even back to exchanges.
594 BTC Drained in the First Wave, More in Subsequent Ones
The incident was not a disorganized event but a cold and meticulously automated operation. The attackers did not have to break the network in real time: they had already precomputed the seed phrases offline through brute force and only waited for the exact moment to issue the withdrawal orders.
|
Attack Phase |
Time Window / Blocks |
Operation Details |
| Initial burst | Three consecutive blocks | 1,324 movements were executed across 500 transactions to drain the first singlesig wallets with more than 0.15 BTC. |
| First wave | Six blocks (~41 minutes) | The automated script completed the sweep of nearly 1,200 addresses, draining an initial 594.5 BTC. |
| Fees and commissions | Entire first wave | Each transaction paid a fixed high fee of 30 sat/vB to ensure priority and immediate inclusion in the blockchain. |
| Escalation and waves 2 and 3 | Weekend and following days | The attack extended to 3 confirmed waves (and a fourth under analysis), reaching more than 1,596 BTC in total (~$110-130M). |
| Victim profile | Consolidated data | Affected more than 7,300 addresses. 88% of the coins had been inactive for >1 year. The median loss was 1.022 BTC per user. |
| Money laundering | Subsequent phase | The drained funds were consolidated into main addresses and began to be moved through mixers to hide their trail. |
Reports from Galaxy Research confirmed the severity of the impact. Furthermore, although it was initially thought that the flaw only affected the Mk3 model, Coinkite admitted that the entropy failure reached the Mk4, Mk5, and Q models, which explains the enormous number of vulnerable wallets exposed in this attack.
A Firmware Change in 2021 Broke the Device’s Randomness
We arrive at the crux of the matter: how was it possible to guess the private keys of thousands of users? In cryptography, the key to everything is entropy (the degree of pure randomness). To create a truly secure seed phrase, a device must use a hardware random number generator (TRNG).
However, in March 2021, with the firmware update 4.0.1 for the Coldcard Mk3 model, a critical technical error occurred:
- The device stopped using the physical randomness generator and began relying on a software substitute (PRNG).
- This software generated randomness based on fixed data such as the chip’s serial number and internal clock registers. None of these values are secret or unpredictable.
- Instead of generating the standard 128 bits of entropy, the randomness dropped to only 40 bits.
Why Did It Take Five Years to Discover?
The software substitute simulated working correctly and passed the random-appearance checks, making it difficult to detect in standard audits.
Security firms like Wizardsardine confirmed that even the Mk4 models shipped in March 2022 already came out of the factory that way. The vulnerability remained hidden from 2021, allowing the problem to grow in silence until it exploded massively in July 2026.
Coinkite’s Response and the Reaction from the Rest of the Sector
Given the magnitude of the incident, the response from Coinkite and the ecosystem focused on mitigating the damage and pointing out the lessons learned for the future of the industry. The manufacturer’s official recommendation is urgent but clear. However, there is a critical warning that every user must understand:
“Updating the firmware DOES NOT fix already vulnerable wallets. If the seed phrase was created with the faulty code, that key remains mathematically weak even if you update the device.”
The priority actions required are:
- Update the firmware:Download the latest official clean version released by Coinkite to prevent the flaw from affecting future operations.
- Create a new seed phrase:Generate a completely new wallet using additional entropy methods (such as at least 50 physical dice rolls or a robust BIP-39 passphrase).
- Migrate funds immediately:Transfer all balances from the old wallet to the newly generated address.
Users who set up their wallets with multisignature schemes (multisig), those who used a strong passphrase, or those who contributed manual randomness with dice at setup were not exposed to the attack.
The Sector’s Reaction and the Audit Problem
For victims who already suffered the fund drain, the recovery margin is very limited: although transactions on the blockchain are traceable and the use of coin mixers is observed, the return depends on the attackers attempting to liquidate the funds on platforms with KYC controls.
What This Case Reveals About Hardware Wallets
This incident debunks a dangerous myth: believing that security is a product you buy and that’s it. Buying a hardware wallet reduces the attack surface but does not eliminate the need for vigilance. Every device is a box that promises to generate genuine randomness, and almost no user has the knowledge to audit whether that promise is fulfilled.
The lesson here is not that cold wallets are useless, but that self-custody is a continuous process, not a static object. The Bitcoin network is as solid as its cryptographic design, but every external link can become an entry point if we place blind trust in it.
Finally, this is not a Bitcoin failure. The network functioned exactly as it should. It was an error in how a device generated the keys. Bitcoin is solid; what needs to be cared for is how we custody it. And that, fortunately, is in our hands.
FAQ
- Was the Bitcoin network attacked or hacked? The Bitcoin network did not suffer any problem. The flaw occurred exclusively in the local software of the Coldcard wallets.
- Why was it so easy to guess users’ keys?A firmware error disabled the physical randomness generator. This made the seed phrases created mathematically predictable.
- If I update my Coldcard’s firmware, are my Bitcoins protected? The update does not fix keys created in the past. You must generate a completely new wallet and move your funds immediately.
- Who was saved from this vulnerability?Those who added manual randomness with dice when setting up the seed, used a strong BIP-39 passphrase, or used multisignature schemes.