Hack Coldcard: NVK invokes 15 audits, Luke Dashjr speaks of lies

Fifteen security reviews, and not a single alarm. Three months after approximately 1,800 BTC were siphoned from Coldcard wallets, the head of Coinkite reopens the case on X to defend his past audits and points to artificial intelligence as the only entity capable of spotting the flaw. Several Bitcoin developers respond that they had warned him.

If you created your seed on a Coldcard before the summer, this dispute changes nothing for you. You still need to migrate to a new recovery phrase.

Key Points:

• On November 9, NVK listed at least fifteen external reviews that missed the flaw.
• According to him, only a recent AI model could uncover this bug slipped into the code in March 2021.
• Luke Dashjr claims to have privately communicated nine remarks about the random number generator to Coinkite.
• Coinkite acknowledges that no comprehensive independent audit of the corrected firmwares exists to date.
• An update of the device does not save an old seed.

Coldcard: NVK's defense rests on fifteen audits

On November 9, Rodolfo Novak, aka NVK, published a long thread on X. He claims that at least 15 external works of review or security research have focused on Coldcard hardware, its firmware, its tools related to the random number generator, or its publication procedures before the flaw was discovered. Among them are two paid reviews, one for the Mk4 in 2022 and one for the Q in 2024. NVK's thread from November 9. Source: X

He also cites a review conducted by developer Rob Ham using OpenAI's cybersecurity tool. The result was 85 potential issues reported and several fixes made, but no trace of the seed generation flaw. << The disturbing truth is that it took a new AI model to find what humans had not found >>, writes NVK.

The bug dates back to March 2021. To create a seed (the sequence of 12 or 24 words that allows the reconstruction of all the keys of a wallet), the firmware called the MicroPython software generator instead of the intended hardware random chip. Keys that were supposed to be unfindable became guessable through brute force. On July 30, a first sweep emptied 1,196 addresses in 41 minutes.

Coldcard Flaw: Luke Dashjr and James O'Beirne claim to have warned

The counterattack came on the same day. Bitcoin Core developer Luke Dashjr published nine observations on Coldcard's random number generator, which he claims to have initially communicated privately to Coinkite. He targets the initialization of the generator, the error handling of the STM32 chips, and the detection of repeated values. He specifies that he only took a quick look and does not claim that each point is exploitable. On AI, however, he does not mince words: presenting these flaws as discoverable by a brilliant AI and by it alone is, for him, pure and simple lying.

James O'Beirne had already recounted in early August that he had begun to scrutinize the firmware in May 2025. He questioned the libngu library and hardcoded constants in the backup software generator. He was reportedly told that the configuration was correct on physical devices. He admits he did not insist.

Tensions rise quickly. Alex Waltz accuses Coinkite of discouraging the examination of its code, while Shinobi defends NVK by reminding that reviewers had indeed inspected the entropy system without seeing anything. And NVK, for his part, repeats that a private message expressing concern does not equate to a disclosure of a reproducible flaw.

Coldcard Wallet: A postmortem still in preparation

Both camps stumble on the same blind spot. As early as August 3, Nick Percoco, security head at Kraken, explained that auditors checked for the presence of the correct generator in the code, without verifying that the production firmware called it. At the end of August, Trezor added that a paid audit from March 2022, which specifically covered random generation, had missed it.

On Coinkite's side, the security status page updated on October 1 lists four independent checks focused on the corrected firmwares. It specifies that there is no complete independent audit of each binary. The detailed technical postmortem, awaited since the summer? Still << in preparation >>. Coldcard security status, updated on October 1. Source: Coinkite

For you, the instructions have not changed one bit. The recommended versions are 5.6.3 for the Mk4 and Mk5 and 1.5.3Q for the Q, but an update does not migrate a seed. The only recognized exception is a seed created with at least 50 fair and never-exposed rolls of dice, which provides about 128 bits of entropy. For all others, a new recovery phrase on an updated device, then transfer of funds.

A small part of the loot has already changed hands in the right direction. On September 21, whitehats gathered 52.37 BTC linked to the Coldcard hack into a trust in Wyoming tasked with returning them to the victims, which is barely 3% of the approximately 1,800 BTC stolen.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.