Bitcoin Core has changed how it handles security flaws, and the trigger is the speed at which AI tools can now dig up bugs. Maintainers merged a revision to the project’s security policy on Thursday that ends the months-long private embargo for any vulnerability a large language model could easily rediscover.
The old system worked on severity. Each reported flaw was graded, and disclosure was scheduled to match. Medium and high-severity issues were commonly kept confidential until the affected versions reached end of life, which in practice could stretch to close to a year. The logic was that node operators upgrade slowly, so a long quiet period gave them time before technical details went public.
That logic has taken a beating. The new policy lets maintainers flag a bug during triage as “easily discoverable,” defined as something that can be found with minimal effort, such as a simple LLM prompt. Flagged fixes are backported to every supported version of Bitcoin Core and pushed out through point releases as soon as possible, alongside a security advisory that carries some of the details. The policy’s own reasoning is blunt: a long embargo achieves little for these bugs, and it only delays the release of a fixed version.
The change follows a summer of AI-driven security scares across the Bitcoin ecosystem. In July, a volunteer group called the Bitcoin Red Team, made up of 16 developers, pointed AI models at roughly 390 Bitcoin-related projects and came back with 4,962 findings in about a day. Eighty-five were rated critical. In August, Core Lightning’s small development team was hit by a wave of AI-generated vulnerability reports, several of which were genuine, and told node operators to upgrade or disconnect from the network. Coldcard, the hardware wallet maker, said it believed attackers had used AI to find a flaw in its firmware, and the Boltz swap service paused operations after saying attackers were finding bugs faster than it could patch them. A group that included Coinbase, Block, BitGo and Blockstream even asked AI labs for early access to their strongest models, arguing Bitcoin developers shouldn’t be locked out of tools attackers can already reach.
The change has critics. Longtime security researcher Antoine Riard argued Core should keep the discretion to ship a fix with no advisory at all and hold details back case by case. There is also a practical tradeoff: patches may go public with less testing time than before. No vulnerability has been placed in the new category yet, so the policy is untested in the real world.
The most serious flaws sit outside the standard timetable anyway. Anything that could threaten Bitcoin’s monetary rules or transaction integrity would still get an ad hoc response. Everything else now runs on a simple assumption: if a chatbot can find it, an attacker already has.
