A hardware wallet’s offline design protects against attacks that need to reach the device over a network. It does nothing against a flaw baked into the wallet’s seed at the moment of creation
A hardware wallet’s offline design protects against attacks that need to reach the device over a network. It does nothing against a flaw baked into the wallet’s seed at the moment of creation — and that is exactly what happened to owners of the Coldcard Bitcoin wallet in the incident this page examines. Attackers never touched the victims’ devices. They rebuilt candidate seed phrases offline and checked them against public blockchain data, according to The Hacker News, whose report on Block’s research was published 1 August 2026.
Coldcard is a Bitcoin-only hardware wallet made by the Canadian firm Coinkite. Block traced the fault to a production configuration error dating to March 2021: the build defined a macro called MICROPY_HW_ENABLE_RNG as zero, because Coinkite supplies its own hardware random-number-generator wrapper, but a library called libngu checked only whether that macro existed rather than whether it was switched on, per The Hacker News. That bound affected builds to MicroPython’s Yasmarang fallback generator, which was seeded from the chip’s unique ID and timer registers and gathered no further randomness afterward.
Coinkite itself estimates the resulting entropy at roughly 40 bits on the Mk3 model and about 72 bits on the Mk4, Mk5 and Q, against the 128 bits a standard 12-word BIP-39 seed is meant to carry, according to The Hacker News. Block did not publish a single practical brute-force benchmark; it instead set conditional ceilings and cautioned that even the higher one does not translate to genuine 73-bit cryptographic strength, per the same report. No public report reviewed for this page has reconstructed a specific victim’s seed and matched it to a drained address.
This is the point that distinguishes the Coldcard case from the threat model most buyers imagine when they choose a hardware wallet. The device never has to connect to the internet, and it didn’t need to here. The weakness lived in the code that produced the seed the day it was generated. Once an attacker can narrow down the device’s unique ID, timer state and prior random-number calls, Block found that candidate seed streams can be generated and tested entirely offline, with the check performed against blockchain data anyone can see, per The Hacker News. The victim’s device, safe, or safety deposit box was never part of the attack surface.
Galaxy Research mapped a sweep of 1,196 Bitcoin addresses completed in 41 minutes on 30 July 2026, moving 1,082.65 BTC worth about $70.2 million at the time, according to The Hacker News. Galaxy told the outlet it found no other Bitcoin transactions in the prior 30 days sharing the same fee rate and no-change signature, and warned that the pattern identifies the operator rather than proving theft, because a sweep “looks the same as if a coin owner chose to move coins.” Galaxy later identified two further suspected waves, raising its observed total to 1,367.05 BTC, worth about $88.6 million, across 4,585 addresses, per The Hacker News, while cautioning that a later wave should not be assumed to share an operator with the first two, and that it has not computationally confirmed every flagged address actually came from the weak-entropy bug.
The mechanism at fault in Coldcard — insufficient randomness at key generation — is not specific to hardware. The Hacker News reported that Coinspect’s separate research, called Ill Bloom, found a weak-PRNG flaw in older software wallets tied to more than $5 million drained since May across Bitcoin, Ethereum, Tron, Rootstock and Polygon addresses. The two flaws are described by the outlet as unconnected incidents, not one bug, but they show that whether a key is generated inside a dedicated device or inside a phone app, the code that produces the random number is the single point either kind of wallet depends on. A wallet’s exposure to this specific class of attack turns on how that code was built and tested, not on whether the device that later holds the key ever touches a network.
Firmware introduces a second, separate risk that has nothing to do with entropy: it is code the manufacturer controls and periodically updates, and a user generally cannot fully audit a given build before installing it. Cointelegraph, in a summary held for this page, described firmware updates broadly as “the perfect attack vector” for a party seeking covert access to private keys — whether, in the outlet’s words, “a regulatory agency — or worse, a criminal organization” — and cited the 2018 Ledger vulnerability as a historical example. That is a different failure mode from the Coldcard case, which traces to an unintended coding error introduced once in 2021, not a deliberately altered update, and the evidence held here shows no connection between the two. Read together, they illustrate the same underlying limit: a hardware wallet’s offline design protects a key from network-based theft, but it does not remove the user’s dependence on trusting the firmware supply chain — whether that dependence is tested by an honest build mistake, as with Coldcard, or by the update channel itself, as Cointelegraph’s framing describes. A software wallet carries a comparable dependency on its own code and update mechanism, without the added step of an air-gapped device.
Coinkite shipped emergency firmware for every affected model and release track on 31 July 2026, per The Hacker News. Coinkite CEO NVK, in an open letter cited by CoinDesk on the same date, told owners to “move your funds now” using updated guidance. Installing the patch stops new seeds from inheriting the flaw. It does not repair a seed already generated on vulnerable firmware: restoring that old seed onto the patched firmware, or onto another wallet, carries the weakness forward, both outlets reported.
Exposure depends on the firmware version running when a given seed was originally created, not the version installed today. Coinkite lists Mk3 versions 4.0.1 through 4.1.9 as affected, fixed in 4.2.0, and does not name the Mk2; Block’s own account places both the Mk2 and Mk3, versions 4.0.0 through 4.1.9, on the vulnerable path, according to The Hacker News. For the Mk4 and Mk5, anything before version 5.6.0 is affected; for the Q, anything before 1.5.0Q; edge builds before 6.6.0X (Mk4/Mk5) or 6.6.0QX (Q) are also affected, per the same report.
Coinkite says a seed generated using at least 50 fair, independent, private dice rolls is not at risk from this bug on its own; where the number or privacy of those rolls is uncertain, it tells users to migrate anyway. A strong, unique BIP-39 passphrase creates a separate wallet the seed words alone cannot reach — but Coinkite still recommends replacing the underlying seed. Multisig setups are only protected where the signing quorum is not built entirely from affected devices. Coinkite’s TAPSIGNER, OPENDIME and SATSCARD products use different codebases and are not affected, according to The Hacker News.
Because Galaxy Research’s on-chain tracing was ongoing, outlets reporting on different dates published different totals for the same unfolding incident. None of these figures is an audited or closed loss total.
Date reportedFigureSource31 July 2026$38 millionCoinDesk1 August 2026 (initial sweep)1,082.65 BTC (~$70.2 million)The Hacker News, citing Galaxy ResearchUndated update within the same report1,367.05 BTC (~$88.6 million), 4,585 addressesThe Hacker News, citing Galaxy ResearchNot stated in the evidence held for this pageover $100 millionBloomberg4 August 2026 (“as of Tuesday”)around $130 millionTechCrunch, citing Galaxy Research; verbally corroborated as roughly correct by Elliptic co-founder and chief scientist Tom RobinsonFor scale, TechCrunch reported, citing blockchain-monitoring firm TRM Labs, that more than 200 hacks targeting cryptocurrency companies had occurred so far in 2026 across the industry, with total losses over $950 million — a broader figure covering many incidents, not Coldcard alone. One individual, Jonathan Goodman, told TechCrunch on X that he lost $1.6 million from a Coldcard wallet despite never sharing his seed phrase or connecting his devices to the internet; that figure is self-reported and unverified in the evidence this page draws on.
This page cannot state a final loss total. Every figure above is a dated snapshot from an on-chain investigation that Galaxy Research described as still ongoing, not a closed or audited number, and the outlets cited gave different amounts on different dates for what may be the same evolving event. The Bloomberg figure of over $100 million carries no publication date in the evidence held for this page, so it cannot be placed precisely on the timeline against the other four figures beyond noting it is one of five estimates in circulation.
It cannot verify Jonathan Goodman’s individual loss claim, which comes from a social media post with no independent or on-chain confirmation in the sources held here. It cannot confirm whether the software-wallet flaw Coinspect calls Ill Bloom is technically related to the Coldcard bug or simply a separate case of the same category of mistake — The Hacker News treats them as distinct incidents without explaining any link. The Cointelegraph material on firmware as a general attack vector is held only as a summary and a single verbatim excerpt, describes a different risk (a compromised update) rather than the entropy bug at issue in Coldcard, and is not corroborated by any other outlet in this pack. No outlet cited here has named the attacker or attackers, and Galaxy Research explicitly warned that a later wave of sweeps should not be assumed to share an operator with earlier ones. This page also holds no primary document — no Coinkite advisory text, Block research paper or Galaxy Research report itself — only news coverage that summarizes them, so the entropy figures and vulnerable-firmware ranges above are one step removed from the original technical source. Finally, this page cannot say whether hardware wallets as a category are more or less exposed to weak-entropy bugs than software wallets; it can only show that this specific failure mode — insufficient randomness at the moment a key is created — has now been documented in a leading product from each category.
Every fact above is attributed to one of these reports. Where they disagree, the article says so.
The post Coldcard’s offline design didn’t stop the theft: what the flaw actually was appeared first on TheCoinrise.com.
