Coldcard Exploit: $130M Lost, 15 Attackers
Coinkite’s Coldcard hardware wallet exploit has now drained an estimated $130 million across at least four attack waves, with Galaxy Research confirming at least 15 separate attackers involved. The breach, traced to a firmware entropy flaw introduced in March 2021, remains active as of Tuesday, and Coldcard has told all users to treat fund migration as urgent. A secondary phishing wave is now compounding the damage, with fraudulent emails directing holders to a cloned Coldcard site that installs remote-access software.
Fifteen Attackers, Four Waves, One Overlooked Bug
Galaxy Research head of research Alex Thorn confirmed Tuesday that victim reports have allowed analysts to identify attackers who would otherwise have gone undetected. “Due to one single victim’s report of less than 1 BTC stolen, we identified a new attack with 12 BTC siphoned from 126 addresses,” Thorn wrote on X. Galaxy counts roughly 2,055 BTC stolen across confirmed and suspected activity, with approximately 90% of stolen funds still sitting unmoved in attacker-controlled wallets. Galaxy has shared all confirmed attacker addresses with US law enforcement agencies, crypto exchanges, and blockchain investigation firms.
The root cause is a single faulty commit dated March 1, 2021, which swapped a call to the device’s hardware random number generator for MicroPython’s software randomizer. Coinkite has acknowledged that affected seeds carry roughly 72 bits of entropy rather than the 128 bits a standard 12-word seed provides, making offline regeneration of private keys computationally feasible. Firmware version 4.0.0 shipped with the flaw on March 17, 2021, and the defect went undiscovered for more than five years while Coldcard’s popularity grew. Devices set up with the dice-roll option using at least 50 physical rolls, as well as wallets protected by a strong BIP-39 passphrase or genuine multisig arrangements, are considered unaffected.
Bitcoin developer James O’Beirne’s August 4 analysis adds a accountability dimension that Coinkite has not yet addressed. Using GPG commit signatures, O’Beirne argues that the pseudonymous account “Switck,” which published the faulty libngu library, is cryptographically the same person as Coinkite co-founder and CTO Peter Gray. O’Beirne states he flagged the entropy risk to Coinkite in May 2025 and was brushed off. The analysis linking Gray to the pseudonymous account rests on 58 commits signed by Gray’s personal key appearing under the Switck identity in the same repository. Coinkite has not responded to the identity claim.
Bitcoin BTC$64,062.00▲0.60%
Ethereum ETH$1,864.27▲0.10%
Tether USDT$0.9993▲0.00%
BNB BNB$598.03▲1.30%
USDC USDC$0.9997▲0.00%
XRP XRP$1.07▼0.90%
Solana SOL$73.58▼0.10%
TRON TRX$0.3263▼0.70%
Figure Heloc FIGR_HELOC$1.00▼0.30%
Hyperliquid HYPE$55.94▲3.20%Phishing Layer and the Exchange Reversal
Hardware wallet firms warned Tuesday of a coordinated phishing campaign exploiting the panic. Fake “coordinated hardware audit” emails are steering Coldcard holders toward a cloned site that installs remote-access software, meaning users fleeing the firmware flaw face a second, independent threat vector if they act on unsolicited communications. Coldcard’s official guidance remains to follow the model-specific advisory, generate a new seed on updated firmware, and migrate funds carefully.
The incident has reversed a behavioral pattern established after the FTX collapse. OKX Chief Compliance Officer Jonathan Brockmeier said the exchange is “seeing record levels of inflows now to centralized exchanges post-Coldcard,” describing it as “the flip side of FTX.” On-chain data corroborates this: Glassnode recorded roughly 980,000 active addresses on July 31, the highest reading since December 2024, driven almost entirely by sending addresses rather than receiving ones. At time of writing, active addresses over 24 hours stand at 523,906, well below that panic peak, suggesting the acute migration wave has partially subsided. The network’s hash rate holds at 941.3 EH/s with 88,921 blocks remaining to the next halving, indicating the protocol itself is unaffected by the wallet-layer failure.
What the evidence establishes, taken together, is a foreseeable failure that was reported internally, not acted upon, and then discovered independently by at least 15 actors within days of each other. That sequence is not a coincidence of timing. It is what happens when a known risk stays unpatched long enough for the information to diffuse.
