A security failure in COLDCARD firmware has challenged one of the strongest assumptions in Bitcoin self-custody: that keeping private keys offline is enough to protect long-term holdings. On July 30, 2026, Coinkite, the company behind COLDCARD, published a security advisory after researchers identified a flaw in the way certain firmware versions generated wallet randomness. Block’s Bitcoin Engineering and Security team published a technical analysis the same day and said active exploitation was already under way.
The incident does not indicate that Bitcoin’s blockchain or core cryptography was broken. The problem was narrower but still severe: some COLDCARD devices could generate wallet seeds from substantially weaker entropy than intended. If an attacker can narrow the possible seed space enough, the attacker may be able to reconstruct a wallet’s private keys offline and move the funds without touching the physical hardware wallet.
What happened
According to Block’s July 30 technical report, a firmware integration error caused a random-number function used by COLDCARD to resolve to MicroPython’s deterministic Yasmarang software fallback instead of the intended STM32 hardware random-number generator. Coinkite’s own technical backgrounder, also published July 30 and later updated, confirmed that a series of integration bugs prevented the hardware RNG from contributing randomness as intended in affected firmware.
The regression entered COLDCARD wallet seed generation in March 2021. Block traced the vulnerable path to firmware released after a change that moved seed generation to a different software library. Coinkite said Mk2 and Mk3 firmware in the affected range produced an estimated effective search space of about 40 bits under its current assumptions. Later Mk4, Mk5 and Q devices mixed in additional secure-element entropy, but Coinkite estimated their effective search space at roughly 72 bits, still below the intended security target.
Block cautioned that these figures do not mean every attacker can instantly recover every affected seed. Real-world exploitability depends on variables including device identifiers, timing state, prior RNG calls and derivation cost. Coinkite likewise described parts of its entropy estimates as preliminary. That uncertainty matters because public estimates of affected wallets and stolen bitcoin have continued to change as researchers analyze blockchain activity.
Which users are affected
Coinkite’s security advisory says funds may be at risk when the seed was generated on Mk2 or Mk3 firmware 4.0.1 through 4.1.9, or on Mk4, Mk5 and Q devices before their fixed firmware versions, unless sufficient independent dice entropy was added during seed creation. The company says seeds created with at least 50 fair, independent and private dice rolls are not considered exposed to this RNG issue alone.
Coinkite also says a strong, unique BIP-39 passphrase adds an independent barrier, but it does not repair a weak seed. A short, common, patterned or reused passphrase may itself be guessable. The company therefore advises affected users to replace the seed even when a strong passphrase reduces immediate exposure.
The fixed versions listed by Coinkite are 4.2.0 or later for Mk2 and Mk3, 5.6.0 or later for standard Mk4 and Mk5 firmware, 1.5.0Q or later for standard Q firmware, and corresponding 6.6.0X or 6.6.0QX releases for the Edge track. Coinkite says TAPSIGNER, OPENDIME and SATSCARD are not affected because they use different codebases.
Why a firmware update is not enough
The most important operational point for investors is that updating firmware changes future seed generation; it does not change a seed that already exists. A seed phrase is the root secret from which wallet keys and addresses are derived. If that root secret was created from weak randomness, reinstalling software or restoring the same seed on another hardware wallet preserves the same vulnerable key material.
Coinkite’s migration guidance therefore requires affected users to install fixed firmware, generate a completely new seed, verify the backup and wallet fingerprint, test a receive address with a small transaction, and then migrate the remaining funds. The company specifically warns users not to generate a replacement seed on an affected device before installing the corrected firmware.
What this changes for Bitcoin investors
The incident shifts the self-custody discussion from a simple question—“Is the wallet offline?”—to a broader question: “How was the key created, verified and protected?” Air-gapping can reduce exposure to internet-connected malware, but it cannot compensate for a private key that was weak from the moment it was generated.
For investors, the practical lesson is diversification of failure modes rather than blind trust in a single security feature. A hardware wallet can still be useful, but its security depends on firmware, random-number generation, supply-chain integrity, backup practices and the user’s operating procedure. More complex custody structures such as multisignature can reduce dependence on one key, but only when the signing threshold cannot be satisfied entirely by keys affected by the same vulnerability. Complexity also introduces its own risks, including configuration mistakes, lost backups and recovery failures.
The event also shows why open-source code is not equivalent to continuously audited code. COLDCARD firmware was publicly available, yet the vulnerable integration path remained unnoticed for years. Block’s report noted that the intended hardware RNG code existed in the firmware while the seed-generation path resolved to the wrong implementation, illustrating how a security failure can survive code visibility if reviewers do not trace the complete execution path.
What remains uncertain
The full financial impact remains unsettled. Blockchain researchers and media reports have published different estimates for stolen bitcoin and affected addresses, and Block explicitly stated that its early report was based on current understanding rather than full empirical exploit testing. Coinkite said its investigation was ongoing and that some technical estimates could change.
Attribution is also unresolved. A cluster of suspicious transactions can show that coins moved in similar patterns, but it does not automatically prove that every address was generated by affected COLDCARD firmware or that one attacker controlled every theft. Investors should therefore distinguish confirmed technical vulnerability from evolving estimates about victim count, total losses and attacker identity.
The broader custody lesson
The Coldcard incident does not establish that self-custody is inherently unsafe, nor does it prove that exchange custody is safer. Each model concentrates different risks. Self-custody transfers responsibility for key generation, backups and transaction signing to the holder, while third-party custody introduces counterparty, operational and access risks.
For beginner and intermediate traders, the main lesson is that “cold storage” is a description of how keys are kept, not a guarantee of security. Robust custody requires understanding where entropy comes from, keeping recovery material private, verifying firmware and addresses, and planning for device failure or software defects. The Coldcard vulnerability is therefore significant not because it broke Bitcoin, but because it exposed how a failure at the key-generation layer can bypass many of the protections investors normally associate with an offline wallet.
This article is for informational purposes only. It is not financial advice, a trading signal or a recommendation to buy, sell or hold Bitcoin or any other asset.