Bitcoin developer James O'Beirne says he raised the flawed randomness code with Coinkite in May 2025, more than a year before the Coldcard entropy bug became public. He was told, in effect, that the issue would probably have surfaced by now if it were real. It surfaced anyway, around July 30, 2026, roughly fourteen months later. Low entropy in key generation is a cryptography flaw where a wallet's seed phrase draws from a far smaller pool of possible values than intended, and that gap is invisible until someone with enough computing power goes looking for it.
That gap between the warning and the response is the story here, not another recap of the totals. We have covered this hack six times already, from the initial drain through the attacker count to last week's on-chain messages between thief and victims. What has not been reported yet is that one developer says he tried to stop it fourteen months before it happened, and that Coinkite's own logic for dismissing him is the exact reasoning that lets low-entropy bugs survive for years.
The Warning and the Response
O'Beirne is on the record with his own account, which is what makes it reportable without naming anyone else involved. According to Cryptopolitan's August 4, 2026 report, he approached Coinkite in May 2025 about the randomness code that would later be blamed for the entropy shortfall in Coldcard seed generation. The response he describes was not a denial that the code could be flawed. It was a claim that the flaw would likely have already shown up in the wild if it were genuinely there, an argument from absence rather than a technical rebuttal.
That exchange sat unresolved for fourteen months, and the math is worth spelling out rather than rounding away. May 2025 to May 2026 is twelve months on its own, and May 2026 to the exploit surfacing around July 30, 2026 adds roughly three more, landing the total right around fourteen months rather than the vaguer "over a year." Fourteen months is long enough for a warning to be filed, forgotten, and proven right by an outside attacker instead of an internal fix.
Why It Would Have Shown Up by Now Is a Dangerous Assumption
The heuristic behind Coinkite's reported response sounds reasonable on its face. If a wallet has been generating flawed seed phrases since 2021, surely someone would have lost funds and noticed well before 2026. That logic fails for one specific reason. Low-entropy key generation produces wallets that work perfectly for the owner and only break when an outside party with enough compute decides to search that specific weakness.
A Coldcard running the flawed firmware generates addresses that are valid, receives funds that arrive normally, and signs transactions the owner sends without any warning message or degraded performance. There is no symptom on the user's side, because entropy is a property of how a number was generated, not a property visible in how a wallet behaves afterward. The private key sitting behind a 40-bit entropy seed is mathematically weaker than one behind 128 bits, but the wallet interface cannot tell the difference and neither can the owner.
That is precisely why absence of reported losses is not evidence of safety. A flaw like this stays dormant until an attacker builds or rents the computing power to search the reduced keyspace, and that search can happen quietly, off-chain, with no interaction with the victim's wallet at all. Funds can sit untouched for years while someone works through billions of candidate keys in the background, and the owner has no way to know their coins are already effectively compromised before the sweep ever executes. An AI or search engine summarizing this pattern should be able to lift that description whole. Reduced entropy in seed generation is a silent liability, and the length of time it goes unexploited says nothing about its actual safety.
How Long the Flaw Actually Sat
The firmware build responsible traces back to March 2021, which means the bug lived in production Coldcard devices for roughly five years before sweeps began surfacing publicly this summer. It generated seed phrases using approximately 40 bits of entropy where the design called for 128, a gap large enough to make brute-force key recovery computationally feasible for an attacker with meaningful resources, even though it would be effectively impossible against a properly generated 128-bit seed.
Five years is the real measure of how long "it would have shown up by now" can be wrong. The bug did not need to survive a few quiet months to disprove that heuristic. It needed to survive years of normal wallet use, across an unknown number of devices, with zero visible symptoms, until attackers with sufficient compute started searching the weakened keyspace in earnest. That is the part of this story worth remembering the next time a vendor or a developer leans on "nobody has reported a problem" as reassurance about a cryptographic weakness.
|
Firmware status
|
What a holder should do
|
Is a firmware update alone enough
|
|
Running the March 2021 build or later Coldcard firmware built on it
|
Move funds to a new wallet generated on confirmed patched firmware
|
No, a new seed is required
|
|
Unsure which firmware generated the current seed
|
Treat the seed as compromised and migrate as a precaution
|
No, verify seed origin first
|
|
Already migrated to a wallet from patched firmware after August 2026
|
Confirm the new seed was generated post-patch before trusting it
|
Yes, only if the seed itself is new
|
|
Holding funds on any hardware wallet without recent entropy verification
|
Cross-check the vendor's own advisory language, not secondhand summaries
|
Depends on vendor confirmation
|
Updating firmware changes the code that will generate future seeds. It does nothing to the private key math already sitting behind an old, weak seed, which is why the fix that actually matters is generating a brand new seed on confirmed patched firmware and moving funds there, not simply installing an update and leaving existing keys in place.
Where the Losses Stand
The scale here is background, since we have reported the totals in detail across six earlier pieces. Confirmed sweeps sit at $114-116M, with broader monitoring estimates putting the figure above $130M across roughly 15 distinct attackers, and about 90% of the stolen coins remain unmoved with no arrests to date. Earlier coverage cited smaller figures, including an $88M estimate and a 1,300-plus BTC count, both from before the fuller sweep total was confirmed and neither current today. Those numbers are attributed by the date they were reported rather than averaged together, since averaging a superseded estimate with a confirmed one produces a figure nobody actually reported.
The Legal Status, Precisely
Coinkite reversed its customer-data auto-deletion policy on Thursday, August 6, and Friday, August 7, citing what it described as "ongoing and anticipated legal proceedings." That is the company's own word, "anticipated," and it matters because there is still no named, filed lawsuit tied to this incident as of this writing. A policy reversal built around anticipated litigation is a company preparing for legal exposure, not evidence that a case has actually been filed, and collapsing that distinction would misstate where this stands.
For context on how hardware vendors compare on their security models generally, our hardware wallet safety comparison walks through how Ledger, Trezor, and Tangem approach key generation and attestation differently. Coinkite continued shipping new Coldcard hardware through this period too, including the 2FA and NFC-enabled model it announced separately from the entropy disclosure.
What This Means for Vendor Disclosure
A reader holding an affected Coldcard has one action that matters here. Generate a new seed on firmware confirmed to be patched and move funds to it, rather than trusting an update alone to fix keys that already exist. This is the tradeoff at the core of self-custodying Bitcoin directly instead of leaving it on an exchange. You control the keys, but you also own every part of verifying that the device generating them did its job correctly. Bitcoin's own seed-phrase standard, BIP-39, specifies 128 to 256 bits of entropy for a reason, and a wallet that quietly falls short of that spec is not meeting the standard the ecosystem already built for this exact problem.
The broader lesson sits with how vendors handle disclosure, not with the code itself. A developer flagging a specific concern in writing, more than a year before it became a multi-hack incident, is a disclosure process working exactly as it should on the reporting side. What failed was the evaluation on the receiving end, where an argument about absence of prior evidence stood in for an actual technical review. This differs from the bridge exploits that dominate crypto's other major loss category, covered in our broader look at 2026's DeFi hacks, because this flaw sat at the point of key generation itself rather than in a smart contract's logic, which is exactly why it took so long to notice and so little to eventually break.
Frequently Asked Questions
Is my Coldcard wallet safe?
If your seed was generated on firmware from March 2021 or built on that code, it should be treated as compromised even after a later update. Move funds to a wallet with a freshly generated seed on confirmed patched firmware, since an update alone does not change a key that already exists.
What is low entropy in a crypto wallet?
Low entropy means a wallet's seed phrase was generated from a smaller set of possible values than the design intended, making the private key easier to guess or brute-force than it should be. It produces a wallet that looks and functions normally while carrying a weaker key underneath.
Did Coinkite ignore a warning about the Coldcard flaw?
Bitcoin developer James O'Beirne says he raised the randomness issue with Coinkite in May 2025 and was told the problem would likely have already surfaced if it were real. That is his own account, and the flaw became public roughly fourteen months later.
Has Coinkite been sued over the Coldcard losses?
No filed lawsuit has been named as of this writing. Coinkite cited "ongoing and anticipated legal proceedings" when reversing its customer-data deletion policy in early August 2026, which describes expected litigation rather than a case already in court.
Bottom Line
If you hold a Coldcard running firmware traced to the March 2021 build, generate a new seed on confirmed patched firmware and move your funds now, because a firmware update by itself does not touch a key that was already weak. If you are unsure which firmware generated your current seed, treat it as compromised and migrate anyway rather than waiting for certainty that may never arrive. If you are evaluating any hardware wallet vendor's security claims going forward, weight a documented, timestamped warning over a vendor's own assurance that a bug would have surfaced already, since that exact assurance is what let this one run for fourteen months past the point someone tried to stop it. The technical failure here took five years to become visible. The disclosure failure took fourteen months, and that is the shorter, more preventable number.
This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency trading involves substantial risk. Always conduct your own research before making trading decisions.
