At least 15 different attackers have exploited the Coldard hardware wallet vulnerability, according to Galaxy Digital head of research Alex Thorn, after new victim reports revealed previously unidentified thefts.
Thorn said the reports allowed Galaxy researchers to connect additional Bitcoin addresses to the incident. In one case, a victim who lost less than 1 Bitcoin helped researchers identify an attacker that siphoned 12 BTC from 126 addresses.
Unlike a centralized exchange breach, the thefts involved funds moving from numerous independently controlled wallets to multiple attacker addresses. Researchers may therefore be unable to identify some thefts unless victims disclose their wallet addresses and transactions.
Galaxy Research estimates that approximately $100 million in Bitcoin was stolen across three confirmed attack waves. Researchers have also identified a suspected fourth wave that could increase total losses to around $130 million.
The 15 attacker clusters do not necessarily represent 15 known individuals. Some addresses could be controlled by the same person or group, while other attackers may remain undiscovered.
Coldcard Bug Weakened Wallet Seed Generation
The vulnerability affected the process used by some Coldcard devices to generate recovery phrases.
A firmware integration error reportedly caused affected wallets to rely on a deterministic pseudorandom number generator instead of the intended hardware-based random-number generator. This reduced the effective randomness protecting some private keys and made it more practical for attackers to search for valid seed phrases.
Castle Labs co-founder Francesco said affected Coldcard wallets may have generated private keys with roughly 40 bits of entropy. A standard 12-word recovery phrase normally provides 128 bits.
Example of a 12 word recovery phrase
This did not mean attackers could directly extract keys from an air-gapped device. Instead, they could recreate possible recovery phrases and monitor the associated Bitcoin addresses for funds.
Users with potentially affected wallets have been advised to update their firmware, generate an entirely new seed phrase and transfer their Bitcoin to addresses controlled by the new wallet. Merely importing an affected recovery phrase into another device does not resolve the underlying weakness.
Could AI Have Prevented the Coldcard Exploit?
The incident has also prompted debate over whether AI-assisted security testing could have discovered the flaw sooner.
Dragonfly managing partner Haseeb Qureshi argued that approximately “$2 of AI hardening” might have prevented the losses. He referenced experiments claiming that AI models rediscovered the vulnerability within minutes, including a test in which the open-source GLM 5.2 model reportedly found it in about 20 minutes without web access.
However, critics questioned whether the experiments prove that AI could have found the vulnerability before it became public.
Tokenomist data lead Tatsapat Saerejittima said that some tests were conducted after disclosure and lacked blind testing, documented methodologies and assessments of false positives.
AI may reduce the cost of identifying vulnerabilities, but it cannot replace independent audits and end-to-end testing of how hardware wallets generate and protect private keys.
