Coldcard Wallet Exploit Reignites Debate Over Open Source Licensing

    Thursday, August 20, 2026 at 08:49 PM UTCOriginally reported by Bitcoin Magazine
    Coldcard Wallet Exploit Reignites Debate Over Open Source Licensing cryptocurrency guide featured image

    A hardware wallet exploit that reportedly cost users more than $100 million in bitcoin has revived an old argument in crypto: does publishing code make software safe? Coldcard's firmware is readable by the public, yet a serious flaw sat in it for roughly five years before attackers used it.

    What separates open source from source-available

    The Free Software Foundation defines free software by four freedoms: to run the program for any purpose, to study and change it, to redistribute copies, and to share modified versions. The Open Source Initiative lists ten criteria, and a license must satisfy every one to count as open source.

    Source-available code can be read by anyone, but its license limits what others may do with it. Coldcard's firmware uses MIT terms combined with the Commons Clause, which takes away the right to sell the software. The clause itself says software carrying it is not open source.

    What went wrong with Coldcard

    • The entropy flaw was introduced in a major 2021 rewrite.
    • That rewrite swapped the trezor-crypto library for libngu.
    • libngu had very little outside attention: 7 stars and fewer than 20 forks, against 512 stars and 212 forks for trezor-crypto.

    The source argues that the commercial restriction shrank the pool of people with a reason to review the code. Most users never read the code themselves and assume someone else has.

    Bitcoin Core as a contrast

    Bitcoin Core uses the MIT license, develops in public and relies on review as its main filter. Reviewers signal their view with labels such as Concept ACK, Approach ACK, ACK and NACK. Funding comes from nonprofits and grants, and no special group of developers holds automatic authority.

    AI changes the picture

    After the incident, the Bitcoin Red Team, backed by OpenSats, used AI models to scan hundreds of open-source Bitcoin repositories. It found thousands of issues, including many critical ones, and disclosed them to maintainers before going public. The team said Chinese open-weight models worked more reliably than closed American ones, which often declined security questions.

    AI-generated contributions also burden maintainers, and some projects have limited issue trackers or AI submissions. Meanwhile, AI can quickly de-obfuscate closed code, weakening the idea of security through obscurity.

    Why it matters

    For anyone learning about crypto, the lesson is that visible code is not the same as reviewed code. Open source makes checking possible, but safety depends on whether skilled people have the incentive and time to look. Licenses shape those incentives, and when real money is at stake, a flaw can sit unnoticed for years. Checking how widely a wallet's code is examined, not only whether it can be read, is a sensible habit.