← Coldcard / Coinkite — Firmware Seed Entropy Exploit (Multi-Actor, August 2026)reviewed 2026-10-16 · 35 claims checked
Fact-check findings
What an automated fact-checker found when it re-read Coldcard / Coinkite — Firmware Seed Entropy Exploit (Multi-Actor, August 2026) against the sources the page cites. Only the most recent review is shown.
Read this first
These findings are produced by an automated reviewer, and its results vary between runs: the same page, checked three times on the same day, came back with 15%, 20% and 34% of its claims disputed, mostly because each run extracted a different number of claims. Treat what follows as leads, not rulings.
“Disputed” means the reviewer could not reconcile the claim with the evidence it cited. It does not mean the claim is false. “Unverifiable” means no reachable source settled it either way.
Nothing here changes the page on its own. A proposed correction is applied only after a human moderator approves it; until then the page reads as it did when reviewed.
disputed
1 claimThe reviewer could not reconcile the claim with the evidence it cited. This is a lead, not a ruling that the claim is false.
- #29[disputed][awaiting moderator]in section: Industry Impact and Self-Custody Debate
“Bitcoin commentator Guy Swann characterized the event as 'the worst hit in bitcoin history to the most knowledgeable and properly secured bitcoiners.'”
reviewerBitcoin commentator Guy Swann characterized the event as 'the worst hit in bitcoin history to the most knowledgeable and properly secured bitcoiners.'Presented in quotation marks as verbatim, but the only sourced version of this quote found uses different wording. The substance (Swann calling it the worst hit to well-secured users) is corroborated; the exact transcription is not.Proposed correction (not yet applied)Bitcoin commentator Guy Swann characterized the event as 'the worst hit in bitcoin history to the most savvy and properly secured holders.'
unverifiable
7 claimsNo source the reviewer could reach confirms or contradicts the claim.
- #8[unverifiable][awaiting moderator]in section: Attack Methodology and On-Chain Evidence
“BlockSec and Galaxy Research tracked the sweeps using heuristic address clustering on coldcardwatch.com.”
reviewerBlockSec and Galaxy Research tracked the sweeps using heuristic address clustering on coldcardwatch.com.Could not confirm that BlockSec and Galaxy Research specifically operate or are credited with coldcardwatch.com; the site does not identify its authors. - #18[unverifiable][awaiting moderator]in section: Actors Involved
“73 confirmed victim reports as of early August”
reviewer73 confirmed victim reports as of early August.Specific figure not located in any consulted source; cannot confirm or refute. - #22[unverifiable][awaiting moderator]in section: Coinkite Response and Patch
“Emergency fixed firmware was released on July 31, 2026 at 9:33 AM EDT for all affected model lines: Mk2/Mk3 version 4.2.0, Mk4/Mk5 version 5.6.0, Q version 1.5.0Q, and respective Edge variants 6.6.0X and 6.6.0QX.”
reviewerEmergency fixed firmware was released on July 31, 2026 at 9:33 AM EDT for all affected model lines.The date and version numbers are confirmed, but no source consulted (including Coinkite's own firmware and disclosure pages) gives a specific '9:33 AM EDT' timestamp; this level of precision cannot be verified. - #23[unverifiable][awaiting moderator]in the timeline
“Coinkite CEO Rodolfo Novak (NVK) issues public apology, accepts full corporate responsibility. Emergency patched firmware released at 9:33 AM EDT for all affected model lines (Mk2/Mk3 v4.2.0, Mk4/Mk5 v5.6.0, Q v1.5.0Q, Edge variants). Second confirmed theft wave recorded. CoinDesk reports approximately $38 million stolen in the initial period.”
reviewerEmergency patched firmware released at 9:33 AM EDT (timeline entry).Same unverifiable timestamp as in the Coinkite Response and Patch section; grouped under the same defect. - #25[unverifiable][awaiting moderator]in section: Legal Exposure and Regulatory Context
“Stoltmann Law Offices, a US firm specializing in investment fraud and cybercrime, publicly solicited Coldcard victims to evaluate potential claims and outlined possible legal theories including breach of express warranty, breach of implied warranty, negligence in development practices, negligent misrepresentation, consumer-protection violations, and product-defect theories.”
reviewerStoltmann Law Offices publicly solicited Coldcard victims and outlined legal theories including breach of warranty, negligence, negligent misrepresentation, consumer-protection violations, and product-defect theories.Could not access the cited primary page to confirm; no independent corroboration of this specific firm's claims list was found elsewhere. - #26[unverifiable][awaiting moderator]in section: Legal Exposure and Regulatory Context
“Stoltmann's public statement notes: 'No court has made findings regarding liability for any reported loss.'”
reviewerStoltmann's public statement notes: 'No court has made findings regarding liability for any reported loss.'Could not confirm this exact quotation against the cited source. - #28[unverifiable][awaiting moderator]in section: Legal Exposure and Regulatory Context
“No regulatory action by the SEC, CFTC, or other financial regulators has been reported in connection with the incident as of the investigation date.”
reviewerNo regulatory action by the SEC, CFTC, or other financial regulators has been reported in connection with the incident.A negative/absence claim; no search turned up any SEC/CFTC action, which is consistent with the page, but none of the cited sources explicitly state that no regulatory action has occurred, so this cannot be affirmatively confirmed as sourced.
partially supported
5 claimsThe cited evidence supports part of the claim but not all of it.
- #14[partially supported][awaiting moderator]in section: Financial Impact and Scale
“Fox Business subsequently reported $89 million across 1,200 addresses.”
reviewerFox Business reported $89 million in losses across 1,200 addresses.The headline/URL do pair '$89M' with '1,200 addresses,' so the page's phrasing is not fabricated, but it implies a single matched figure where the article actually reports the address count for the first wave and the dollar figure as a later cumulative update. - #15[partially supported][awaiting moderator]in section: Actors Involved
“Galaxy Research's Alex Thorn stated that at least 15 independent attackers are now exploiting the vulnerability, characterizing the situation as 'an open scramble of independent actors and imitators' rather than a coordinated single-group operation.”
reviewerGalaxy Research's Alex Thorn said at least 15 independent attackers are exploiting the vulnerability, describing it as 'an open scramble of independent actors and imitators.'The 15-attacker figure is well corroborated across outlets; the specific quoted phrase 'an open scramble of independent actors and imitators' could not be located in any source consulted and may be an invented or paraphrased quotation. - #17[partially supported][awaiting moderator]in section: Actors Involved
“On the researcher side, Block Engineering published the technical entropy analysis; Galaxy Research mapped the sweep patterns and coordinated victim reconciliation (73 confirmed victim reports as of early August).”
reviewer'Block Engineering' published the technical entropy analysis.Substance confirmed; naming is imprecise and could mislead readers into thinking 'Block Engineering' is a distinct named research firm. - #27[partially supported][awaiting moderator]in section: Legal Exposure and Regulatory Context
“News.Bitcoin.com reported that victims are planning class-action lawsuits and that legal experts are 'sharply divided' on Coinkite's liability.”
reviewerNews.Bitcoin.com reported that victims are planning class-action lawsuits and that legal experts are 'sharply divided' on Coinkite's liability.The 'sharply divided' quote is verbatim; the characterization of victims 'planning' lawsuits overstates the article's more tentative framing. - #35[partially supported][awaiting moderator]in the summary
“As of August 10, 2026, losses exceed 2,055 BTC (approximately $130 million USD) across more than 7,700 addresses, making this the largest hardware wallet exploit on record.”
reviewerAs of August 10, 2026, losses exceed 2,055 BTC (~$130 million) across more than 7,700 addresses.The BTC/dollar figures (2,055 BTC, $130M) are well corroborated, but 'more than 7,700 addresses' exceeds the best-sourced figure of ~7,300 found in any consulted reporting as of the relevant dates.
confirmed
22 claimsThe cited evidence supports the claim as written.
- #1[confirmed][no action needed]in section: Vulnerability Overview
“The build guard used the expression `#ifndef MICROPY_HW_ENABLE_RNG`, which evaluates to true when the macro is defined as `0` — a state that Coldcard's production board configuration set explicitly to signal use of its own hardware-RNG wrapper.”
reviewerThe build guard used #ifndef MICROPY_HW_ENABLE_RNG, which evaluates true when the macro is defined as 0, causing silent fallback to software RNG.Independently corroborated by Block's engineering blog and secondary technical coverage (akitaonrails) describing the identical mechanism. - #2[confirmed][no action needed]in section: Vulnerability Overview
“On Mk2/Mk3 hardware, no secure-element entropy was mixed into seed generation at all, yielding an estimated 40 bits of effective entropy.”
reviewerEffective entropy fell to ~40 bits on Mk2/Mk3 and ~72 bits on Mk4/Mk5/Q against a 128-bit target.Matches Coinkite's own stated estimate, though Block's independent analysis uses a stricter 2^32-candidate framing that is not interchangeable — a nuance the page does not flag. - #3[confirmed][no action needed]in section: Vulnerability Overview
“Coinkite's official disclosure page classifies the root cause as 'a build-integration and symbol-resolution defect, not an intentional runtime fallback.'”
reviewerCoinkite's official disclosure classifies the root cause as 'a build-integration and symbol-resolution defect, not an intentional runtime fallback.'Verbatim quote confirmed on coinkite.com/historical-disclosures, which is cited among this section's sources. - #4[confirmed][no action needed]in section: Affected Firmware Versions and Models
“Coinkite confirmed the following firmware ranges as vulnerable. Mk2 and Mk3 devices running firmware 4.0.1 through 4.1.9 received no hardware-RNG contribution to seed generation and face the most severe entropy degradation (approximately 40 bits).”
reviewerMk2/Mk3 firmware 4.0.1-4.1.9 is the most severely affected range (~40 bits entropy).Confirmed directly against Coinkite's disclosure history page. - #5[confirmed][no action needed]in section: Affected Firmware Versions and Models
“TAPSIGNER, OPENDIME, and SATSCARD products were confirmed unaffected. Users who supplemented device seed generation with at least 50 independent, private dice rolls are stated by Coinkite to be unaffected by this specific flaw alone.”
reviewerTAPSIGNER, OPENDIME and SATSCARD were confirmed unaffected; 50 dice rolls is Coinkite's stated exception threshold.Confirmed. - #6[confirmed][no action needed]in section: Attack Methodology and On-Chain Evidence
“The initial wave on July 30, 2026 drained approximately 1,082.65 BTC from 1,196 addresses within 41 minutes, suggesting automated tooling.”
reviewerFirst wave on July 30, 2026 drained ~1,082.65 BTC from 1,196 addresses within 41 minutes.Confirmed exactly. - #7[confirmed][no action needed]in section: Attack Methodology and On-Chain Evidence
“On-chain indicators noted by TRM Labs include a single 64.9 BTC deposit to a Wasabi Wallet coinjoin and approximately 200 ETH equivalent sent to Tornado Cash on August 4 — both consistent with early-stage laundering attempts.”
reviewerTRM Labs noted a single 64.9 BTC Wasabi Wallet deposit and ~200 ETH to Tornado Cash on August 4 as early laundering indicators.Confirmed; minor stylistic addition of 'coinjoin' is an accurate description of Wasabi Wallet's function rather than a misstatement. - #9[confirmed][no action needed]in section: Attack Methodology and On-Chain Evidence
“Victims averaged 3.18 years of wallet dormancy before being swept, indicating the primary affected population is long-term Bitcoin holders.”
reviewerVictims averaged 3.18 years of wallet dormancy before being swept.Confirmed via independent secondary coverage of Thorn's statement. - #10[confirmed][no action needed]in section: Financial Impact and Scale
“The Hacker News reported approximately $70.2 million (1,082.65 BTC) confirmed stolen in the initial 41-minute wave on July 30 alone.”
reviewerThe Hacker News reported ~$70.2 million (1,082.65 BTC) stolen in the initial 41-minute wave.Confirmed. - #11[confirmed][no action needed]in section: Financial Impact and Scale
“TRM Labs and Fortune placed losses at approximately $116 million (1,816 BTC across 5,200+ addresses) as of August 3–5.”
reviewerTRM Labs and Fortune placed losses at ~$116 million (1,816 BTC across 5,200+ addresses) as of August 3-5.Confirmed. - #12[confirmed][no action needed]in section: Financial Impact and Scale
“Galaxy Research's August 4 estimate extended the range to 1,596 BTC confirmed across approximately 7,300 addresses, with a possible 2,055 BTC ($130 million) if a fourth suspected wave is fully attributed.”
reviewerGalaxy Research's August 4 estimate: 1,596 BTC confirmed across ~7,300 addresses, up to 2,055 BTC ($130M) if a fourth wave is attributed.Confirmed verbatim against The Block article. - #13[confirmed][no action needed]in section: Financial Impact and Scale
“TRM Labs classifies this as the third-largest crypto hack of 2026 and the largest hardware wallet exploit on record. The year-to-date 2026 total across all crypto incidents exceeds $1.2 billion across 276 incidents per TRM Labs.”
reviewerTRM Labs classifies this as the third-largest crypto hack of 2026; YTD 2026 crypto losses exceed $1.2B across 276 incidents.Confirmed. - #16[confirmed][no action needed]in section: Actors Involved
“TRM Labs notes that the limited layering and mixing activity observed — in contrast to state-sponsored actor patterns such as North Korea's TraderTraitor — suggests opportunistic rather than professional-grade operations.”
reviewerTRM Labs: limited layering/mixing contrasts with state-sponsored patterns like North Korea's TraderTraitor, suggesting opportunistic rather than professional-grade operations.Confirmed. - #19[confirmed][no action needed]in section: Actors Involved
“Ledger CTO Charles Guillemet offered public technical commentary comparing the incident to a similar 2022 Trust Wallet entropy vulnerability discovered by Ledger's Donjon research team.”
reviewerLedger CTO Charles Guillemet compared the incident to a 2022 Trust Wallet entropy vulnerability discovered by Ledger's Donjon team.Confirmed directly from the Ledger blog post. - #20[confirmed][no action needed]in section: Actors Involved
“Security researcher James O'Beirne is cited as publicly warning that single-key Mk3 wallets generated between 2021 and 2023 without passphrases, dice rolls, or multisig should be treated as immediately at risk.”
reviewerJames O'Beirne publicly warned single-key Mk3 wallets generated 2021-2023 without passphrases/dice/multisig should be treated as immediately at risk.Confirmed via independent reporting. - #21[confirmed][no action needed]in section: Actors Involved
“Coinkite's CEO, Rodolfo Novak (publicly known as NVK), issued a public apology on July 31, 2026, accepting full responsibility for the firmware failure. NVK also suggested that AI-assisted code review may have played a role in discovering or accelerating exploitation of the vulnerability.”
reviewerNVK issued a public apology on July 31, 2026, accepting full responsibility, and suggested AI-assisted code review may have played a role.Confirmed. - #24[confirmed][no action needed]in section: Coinkite Response and Patch
“Coinkite's advisory note states: 'The last three days have been some of the hardest in this company's history.'”
reviewerCoinkite's advisory states: 'The last three days have been some of the hardest in this company's history.'Confirmed via secondary reporting quoting the same statement; minor 'some of' phrasing difference across outlets does not change the substance. - #30[confirmed][no action needed]in section: Industry Impact and Self-Custody Debate
“Lorenzo Valente of ARK Invest noted that users had 'traded counterparty risk for software risk, hardware risk, supply-chain risk, phishing risk, backup risk.'”
reviewerLorenzo Valente of ARK Invest said users had 'traded counterparty risk for software risk, hardware risk, supply-chain risk, phishing risk, backup risk.'Confirmed, essentially verbatim. - #31[confirmed][no action needed]in section: Industry Impact and Self-Custody Debate
“Casa CEO Nick Neuman criticized Coinkite's guidance recommending 50 dice rolls as a mitigation, calling it impractical for mainstream adoption.”
reviewerCasa CEO Nick Neuman criticized Coinkite's 50-dice-roll guidance as impractical for mainstream adoption.Confirmed. - #32[confirmed][no action needed]in section: Industry Impact and Self-Custody Debate
“Ledger used the incident to publicly contrast its own security certification approach, noting its use of AIS-31, PTG.2, Common Criteria EAL5+/EAL6+-evaluated entropy sources.”
reviewerLedger contrasted its own security certification (AIS-31, PTG.2, Common Criteria EAL5+/EAL6+ entropy sources).Confirmed directly from the Ledger blog post. - #33[confirmed][no action needed]in section: Industry Impact and Self-Custody Debate
“Bitcoin Magazine alleged that AI-assisted code review tools may have played a role in identifying or exploiting the vulnerability, citing NVK's own public framing, though no confirmed evidence of AI-generated exploit tooling has been published.”
reviewerBitcoin Magazine alleged AI-assisted code review tools may have played a role, citing NVK's framing, with no confirmed evidence of AI-generated exploit tooling.Confirmed, and the page appropriately hedges that no confirmed evidence has been published. - #34[confirmed][no action needed]in the summary
“reducing effective entropy from the intended 128 bits to as low as 40 bits on Mk3 devices and approximately 72 bits on Mk4, Mk5, and Q models”
reviewerEffective entropy was reduced from the intended 128 bits to as low as 40 bits on Mk3 devices and approximately 72 bits on Mk4, Mk5 and Q models.Confirmed, consistent with the detailed section-level finding above.
How this fits together. The reviewer reads the published page and its cited sources and records one finding per claim. A human moderator decides whether each proposed correction is applied; those decisions, and the score changes they cause, appear in the audit log. Earlier review runs are not shown here; only the latest reflects the page as it stands.