---
title: "Three Chains Prepped for Quantum Attacks. A Weak Random Number Broke First."
published: 2026-09-15T02:12:36.268745+00:00
type: state_of_market
scope: crypto_market
canonical: https://www.moonwire.org/insights/state-of-market-2026-08.html
tags: [security, self-custody, hardware-wallets, post-quantum, custody, entropy, coldcard, institutional-custody, stablecoin-rails, etf-flows, monthly, state-of-market, august-2026]
---

# Three Chains Prepped for Quantum Attacks. A Weak Random Number Broke First.

> The one break of crypto's core cryptography in our August coverage was not a quantum computer. It was a five-year-old build flaw that made Coldcard seeds guessable - and unlike the post-quantum upgrades Sui, Starknet and Ethereum advanced in the same fortnight, it came with no route across for anyone who already owned a device. What the voices we track argued about afterwards was not price. It was custody itself - who holds the key, and on whose hardware.

## Key takeaways

- August's failure of key generation was a five-year-old build flaw in a hardware wallet, not a quantum computer: 594 BTC left 500 Coldcards in a 25-minute window, tracing to firmware shipped in March 2021 [[2]](/s/PFZDUAobRLy9a-91slzWKA).
- The patch protected new seeds only - Coldcard's fix "does not restore security to old, vulnerable firmware" [[8]](/s/6RwaZL03Qf-ndVI4ciE4jQ), and Changpeng Zhao noted developers "have no way to reach users on air-gapped devices" [[9]](/s/raXaL-SCRl-qujHLtBHvIQ).
- Sui, Starknet and Ethereum all advanced post-quantum work the same fortnight, with years of lead time: Sui's upgrade runs through existing recovery phrases [[14]](/s/w9qC3lulSWyYReI4bCHiuQ), Starknet tested a post-quantum transfer on mainnet for about $0.06 [[15]](/s/wHW7WNN_QvatnZpjqnDsOQ), and Ethereum put it on a roadmap running to 2029 [[16]](/s/BqLTDaRpT16IFJEUdOgXHg).
- Holders answered by moving rather than selling: active bitcoin addresses hit 0.98 million a day, the highest since December 2024, which Glassnode called an operational security response rather than a change in conviction [[18]](/s/IUVzpIbaRKaIiztaedt6sg), while OKX reported "record levels of inflows" to centralized exchanges [[21]](/s/fEKU11-jSQaw39LR92F32g).
- The price ignored the money again: $211.5 million of spot ETF inflows and a record S&P 500 close left bitcoin near $64,000 [[38]](/s/QI94fWHCRt2Bsji1tIDspw), and perpetual activity on Binance and Bybit fell to its lowest since 2023 [[43]](/s/yaLEVUxvQPejLyzYvzmhpQ).

Crypto's security agenda in August pointed at a computer that does not exist yet. Three chains advanced post-quantum work inside two weeks — and the break that was still emptying wallets as the month opened was a build flaw from 2021 that made bitcoin seeds guessable, inside a device Coin Bureau called the gold standard of [self-custody](https://www.moonwire.org/insights/glossary/self-custody.md) [[1]](/s/n5-edn2UR4SRGH56ERZ-XQ).

The contrast is not only rhetorical. The post-quantum work arrived with years of lead time and, in one case, an explicit route for accounts that already exist. The Coldcard flaw arrived with neither.

## A flaw in how the device made keys

August's first hard number came through CoinDesk, reposted by Changpeng Zhao [[2]](/s/PFZDUAobRLy9a-91slzWKA): 594 BTC — about $38 million — drained from 500 Coldcard hardware wallets in a 25-minute window, "exploiting a flaw in how the device generated private keys," traced to firmware 4.0.0 shipped in March 2021.

The estimate then climbed for a week. Decrypt had it above $70 million on August 1 [[3]](/s/hLaFyuVOSJmGLezYG4xBew), at roughly $88 million on the 2nd [[4]](/s/DNOEJi2bTJKK4XFB2WohTw), and near $114 million on the 3rd alongside a spike in small bitcoin transfers [[5]](/s/X_eDjdBOSPigCYoCmOC5dA). Galaxy Research put 1,596 BTC as confirmed stolen on the 4th, with a potential total of $130 million [[6]](/s/b7FSo4bRSgy-c0zAfleaig), then 1,719 BTC — about $111 million — on the 7th, still expecting losses past $130 million [[7]](/s/Ab9wKduNSSyU4c7747NKsQ). Coin Bureau's breakdown put the effective seed entropy on affected MK2 and MK3 devices at roughly 40 bits and the sweep at between 5,200 and 7,300 addresses by August 3, and said what makes this unlike the [exploits](https://www.moonwire.org/insights/glossary/crypto-exploit.md) that channel usually covers is that, in the cases it reviewed, no victim was phished, no user approved a malicious transaction, and the devices were sitting offline in safes while the coins moved [[1]](/s/n5-edn2UR4SRGH56ERZ-XQ). The drained balances had, on average, gone untouched for more than three years.

Then the part that defines the month. Coldcard destroyed its remaining vulnerable inventory and halted shipments, and its patched firmware "prevents the issue from affecting newly generated seeds but does not restore security to old, vulnerable firmware," The Block reported [[8]](/s/6RwaZL03Qf-ndVI4ciE4jQ). Zhao put the consequence in one line: "devs patching the bug won't fix previously generated wallets. And devs have no way to reach users on air-gapped devices. Your wallet stays open to hackers until you act." He added, in the same post, that he remains a believer in self-custody — "but it puts the burden on you" [[9]](/s/raXaL-SCRl-qujHLtBHvIQ).

Of the failures disclosed across the coverage we tracked in August, this was the one that reached key generation itself. The others landed a layer out: a critical vulnerability in the BTCPay Server payment stack [[10]](/s/-dVzUulGQr6BX6dNMPZC1g), a consensus-rule exploit that sent Ravencoin to an all-time low [[11]](/s/qLIBxty3Roqz3GI98wvWTQ), a large-scale exploit at Harmony that prompted a rollback and exchange alerts [[12]](/s/fTBdrnMyS2KIqOnhAbC_Bw), and a data breach at one of Trezor's shipping providers that exposed customer names and contact details, with Trezor saying its systems and devices remained secure [[13]](/s/Bh4RmHbzQ0OZfO5quTCj0g).

## The upgrade paths were being built for a different attacker

In the same fortnight, Sui introduced quantum-resistant signatures "allowing users to upgrade existing accounts through recovery phrases" [[14]](/s/w9qC3lulSWyYReI4bCHiuQ). Starknet completed a quantum-resistant signature transfer test on mainnet at a fee of about $0.06 [[15]](/s/wHW7WNN_QvatnZpjqnDsOQ). Vitalik Buterin's revised Ethereum "Strawmap" gave first-class attention to strong privacy and aggressive post-quantum scaling — an entirely new addition against his 2023 roadmap — across protocol upgrades running to 2029 [[16]](/s/BqLTDaRpT16IFJEUdOgXHg). StarkWare chief executive Eli Ben-Sasson said privacy and post-quantum security should not be expensive [[17]](/s/ULD4ES3HS7efmiij7ZWBtg).

Read side by side, the asymmetry is the story. A future cryptographic break gets a named threat, a multi-year schedule, a mainnet rehearsal priced in cents, and at least one chain offering existing account holders a way across. The cryptographic break that had already happened — five years old, shipped in a box, sitting in a safe — got a patch that could only protect keys that had not yet been generated.

## What holders did about it

They moved, and the move was measurable. Glassnode [[18]](/s/IUVzpIbaRKaIiztaedt6sg) recorded active bitcoin addresses surging to 0.98 million a day, the highest since December 2024, and drew the distinction carefully: "fear-driven [on-chain activity](https://www.moonwire.org/insights/glossary/on-chain-data.md). Holders migrating seeds and moving funds to alternative [custody](https://www.moonwire.org/insights/glossary/crypto-custody.md) reflects an operational security response, not a change in market conviction." K33, reported by The Block [[19]](/s/7L9XlbLcTeO4-LHNFJYjKA), put bitcoin's seven-day active supply at its highest level of 2026, with more than 890,000 BTC moved in a week. Cointelegraph, citing Santiment, reported 2.27 million new bitcoin wallets and 751,000 active wallets as the Coldcard fallout pushed users to move funds [[20]](/s/SnIQXqmKSmmBITbLwsSq2w).

Some of that traffic went in one direction in particular. OKX [[21]](/s/fEKU11-jSQaw39LR92F32g) said the exploit was driving "record levels of inflows" to centralized exchanges, calling it "the flip side of FTX."

## Where the voices we track split

There was no single verdict, and the disagreement was substantive rather than tonal.

Jameson Lopp [[22]](/s/IMSYq_h1RkO2MUQ4aUoomw) argued the exploit is not the death of bitcoin self-custody and should prompt users to rethink trust assumptions rather than abandon the practice: "Self-custody is for anyone who is willing to take on the responsibility that comes with it."

Zhao took the other side and then qualified it himself. Reposting Willy Woo's figures [[23]](/s/bDuLj06-Sy-RLY2WxWq3GA) — 1.57 million BTC lost via self-custody against 1.51 million lost on exchanges — he wrote that it is "statistically safer to store coins on exchanges than to self custody," then immediately noted that hack data is easier to collect on the exchange side, that self-custody losses often go unreported, and that defunct exchanges drag the comparison down. His conclusion was not a recommendation: "Not saying one is better than another... A balanced approach is probably best."

Bitcoin's official X account [[24]](/s/P8YKlQZkSJ-Hp3V7Ad3w5Q) went the other way, saying true believers have been running cold wallets on general-purpose hardware "for a reason." The analyst XO [[25]](/s/5XAXofAKQEK1_xCJViG95w) treated it as a specification problem rather than a custody-model problem, describing a minimum standard of a 2-of-3 or 3-of-5 multisig using BIP39 with a passphrase carrying at least 128 bits of entropy, and taking care that a seed is not generated solely from a device's own hardware random number generator. Weeks later, well after the news cycle had moved on, Ansem [[26]](/s/iYeNgi4ZTJSVgvI4i0498A) listed the products crypto still needs and put near the top "a hardware wallet that isn't terrible. most get hacked or don't work with the protocols you actually use."

## The institutions spent August building the other kind of key

While holders were rotating seeds, the custody business kept shipping — and the arrivals in this first group each put a regulated intermediary between the holder and the key.

BNY [[27]](/s/WUdrWSCHS8eQAGMFOpW7qw) brought in Galaxy to add digital-asset [staking](https://www.moonwire.org/insights/glossary/staking.md) to its institutional custody platform, combining custody and staking in a single servicing model pending regulatory review. BitGo [[28]](/s/wDFcp3CoRJGwtQ1lVS6H_A) launched Link, connecting institutional client accounts at Coinbase, Kraken and Crypto.com directly to its own platform. SharpLink's $200 million ETH staking allocation through Lido was custodied at Anchorage Digital [[29]](/s/tH6zLlmmSeSLqY0Ea7SjaA).

The settlement layer moved in parallel. Wells Fargo [[30]](/s/tC-UFI8JTpW5WXS798MrfQ) said it will launch tokenized US dollar and sterling deposits on its own blockchain for corporate clients. Circle [[31]](/s/AsJizuCNQM-BvqUFHHiUCA) named 11 founding validators for its Arc chain — including BlackRock, Visa, Mastercard and DTCC — ahead of a September 16 mainnet, with DTC-custodied asset tokenization targeted from the second half of 2027. Mastercard bought BVNK [[32]](/s/K1YYw7NzSxeRIYeZLfxyOA), piloted a [stablecoin](https://www.moonwire.org/insights/glossary/stablecoin.md) payment credential with Borderless.xyz [[33]](/s/_atQ2S_IT4S61-3yjjclkw), and on August 27 added BNB Chain to its Crypto Partner Program for stablecoin payments and on- and off-ramps [[34]](/s/AE2LevrURVSjrak7KuRnMw). Visa Direct tied up with ZeroHash [[35]](/s/c2ENqoV7QoqSL6iBiITzDw), and Western Union launched a Solana-based Stablecard usable at 175 million Visa merchants [[36]](/s/-lxvJN54SBKo1C6hawHu1Q). The [SEC](https://www.moonwire.org/insights/glossary/sec.md) cleared Franklin Templeton's FOBXX on-chain money-market fund [[37]](/s/w7Ty3frJQnWH0sa95UGXrQ).

That is July's pattern continuing rather than a new one: the institutional arrivals keep landing on rails and custody rather than on directional exposure.

## The test we printed last month

July's review closed with a question — would August flows start moving the price again? They did not.

On August 5 The Block reported bitcoin sitting near $64,000 despite $211.5 million of spot BTC [ETF inflows](https://www.moonwire.org/insights/glossary/etf-flows.md) the previous day and a record S&P 500 close, with Bitfinex, Glassnode and Wintermute all reading bottom signals "forming through boredom, not [capitulation](https://www.moonwire.org/insights/glossary/capitulation.md), with no demand engine behind them yet" [[38]](/s/QI94fWHCRt2Bsji1tIDspw). The flows got larger, not smaller — $244.42 million into bitcoin and $60.86 million into ether on August 5 [[39]](/s/OTYaI5pzRPKgx3TeTgmhlA), $854 million into spot bitcoin funds over five days, reported on the 10th [[40]](/s/w4YMu_kmSqeUxdmriZDV0A), in the best inflow week for US spot bitcoin and ether funds since April [[41]](/s/944RBN7CR5mQAMXFCh5SVw) — and the range held. Michael Saylor posted a reading on August 2 putting bitcoin at $63,498 against a 200-week [moving average](https://www.moonwire.org/insights/glossary/moving-average.md) of $63,476 — a 0.0% premium [[42]](/s/EeHA7nVESWSn87a5TrH1dg) — and it was near $64,000 again after July inflation matched forecasts at 3.4% headline and 2.5% core, an in-line print that The Block said "hands crypto no fresh catalyst to break its range" [[43]](/s/yaLEVUxvQPejLyzYvzmhpQ).

The second measure went the same way. CEX futures volume fell to $4 trillion in July, its lowest since December 2023 [[44]](/s/NCiwBuBSS8mhhs42FZnrEw), and K33 put bitcoin perpetual activity on Binance and Bybit at its lowest level since 2023, with spot volumes at their weakest since February 2024 [[43]](/s/yaLEVUxvQPejLyzYvzmhpQ).

So the most consequential movement of bitcoin in August was not bought or sold. It was migrated — by holders changing where a key lives, and by institutions building places for keys to live that a holder never touches.

*This review covers the source coverage we tracked between August 1–13 and August 25–31, 2026.*

---

*MoonWire surfaces and summarizes public crypto news.*

**General information only — not financial advice.** MoonWire Ltd. is not licensed by ASIC and holds no AFSL. This content does not consider your objectives, financial situation, or needs, and is not a recommendation to buy, sell, or hold any asset.

*This analysis is generated by artificial intelligence and may be inaccurate, incomplete, or fabricated. Independently verify before acting.*

© 2026 MoonWire Ltd. · [Terms](https://www.moonwire.org/legal/TOS) · [Risk Disclosure](https://www.moonwire.org/legal/RISK_DISCLOSURE) · [Privacy Policy](https://www.moonwire.org/legal/PRIVACY_POLICY)
