COLDCARD wallet RNG flaw likely linked to $88 million Bitcoin theft

A vulnerability in COLDCARD hardware wallet firmware allowed attackers to steal an estimated $88.6 million in Bitcoin from thousands of wallets. The thefts traced to wallets whose seeds were generated with a flawed random number generator. Because seed material is the root of key derivation, weak randomness at generation time left those wallets recoverable by anyone who could exploit the flaw.

The issue sits in seed generation, not in later transaction signing. Hardware wallets derive private keys from a master seed; if the random number generator that produces that seed is biased or predictable, the seed space collapses and keys can be recomputed offline. Attackers did not need ongoing access to the devices. They needed only knowledge of the flawed RNG behavior and the ability to search the reduced set of possible seeds, then sweep funds once matching addresses appeared on-chain.

For engineers and builders of custody systems, this is a reminder that entropy quality is a hard security boundary. A wallet can implement sound cryptography and still fail if the seed entropy is weak. Seed generation is a one-time operation with permanent consequences: a compromised seed cannot be rotated without moving funds, and users who generated seeds under the flawed firmware remain exposed until they migrate.

Hardware wallet security is often marketed as the gold standard for self-custody, and COLDCARD sits in that market as a cold-storage product aimed at users who want offline key storage. An RNG flaw at this scale undercuts the assumption that air-gapped hardware alone guarantees seed secrecy. Competing designs that separate entropy sources, document randomness testing, or support seed verification give users a concrete way to compare products after incidents like this.

Anyone who generated a COLDCARD seed under the affected firmware should treat that seed as compromised and move funds to a wallet created with a verified entropy source. Builders should require explicit RNG design, testing, and audit evidence for any product that generates seeds, and should plan migration paths when a seed-generation bug is disclosed. Watch for official firmware guidance on which devices and seed-generation paths were affected, and for on-chain analysis that maps how many of the estimated $88.6 million in losses have been recovered or remain at risk.