Bitcoin Core changes its rules in the face of flaws found by AI

Bitcoin Core is changing the way it handles certain vulnerabilities because of artificial intelligence. Developers now recognize that a bug can be so easy to find that a simple prompt sent to an AI model is sometimes enough. In this case, keeping the flaw secret for several months no longer really protects users. Bitcoin Core prefers to patch faster and publish the security advisory at the same time. No new critical Bitcoin bugs are announced. It is defense policy that is evolving.

Bitcoin Core changes its rules in the face of AI

The change was announced by Antoine Poinsot, developer of Bitcoin Core. It adds a new category to the security policy: “easily discoverable” vulnerabilities. The decision comes a few days after the release of Bitcoin Core 32.0, already audited with the help of AI models.

The principle is quite simple. When a flaw falls into this new category, developers will no longer necessarily wait for the next major version of Bitcoin Core. The fix can be backported to all major versions still officially maintained.

Intermediate versions can then be published quickly. The security notice will arrive at the same time. Bitcoin Core notably defines as easily discoverable a vulnerability that can be identified with minimal effort, for example using a simple prompt sent to a large language model.

AI, however, is just one example. A vulnerability that is obvious for other reasons may also fall into this category. The ranking is still decided by the developers when they analyze the problem. This nuance matters: Bitcoin Core does not create a special category called “bugs found by ChatGPT”. Above all, the project adapts its policy to the real level of difficulty necessary to find a flaw. And this level has just fallen.

Long embargoes protected better before LLMs

Traditional vulnerability management relies heavily on temporary secrecy. A researcher discovers a problem. He discreetly communicates it to developers. They prepare a patch and integrate it into a new version without necessarily immediately explaining the risk.

Details come later. This method gives users time to update their software before attackers have a complete user manual.

Bitcoin Core operates with a fairly specific disclosure schedule. A low severity vulnerability may be made public two weeks after the release of the major version containing its fix. For medium or high severity issues, the wait may be much longer. Publication normally occurs two weeks after the last affected version reaches the end of its life.

In some cases this is around a year. Critical bugs follow a different procedure. Bitcoin Core notably places in this category problems that could allow the creation of bitcoins outside of the planned monetary calendar, the theft of BTC at the protocol level or a permanent separation of the network.

These flaws remain exceptional cases. The arrival of LLM especially modifies the reasoning for bugs that are less difficult to reconstruct. The Bitcoin Core code is public. So are changes to the software. An external developer can therefore compare two versions and find out why a few lines have suddenly changed.

Before AI, this operation could require a lot of skill and several hours or days. Now some of the work can be automated. A model can read the change, examine the affected functions, and offer an explanation for the vulnerability being fixed.

Nearly 5,000 Bitcoin bugs reported in 27 hours

Evolution is not theoretical. In early August, the Bitcoin Red Team demonstrated the scale that vulnerability scanning could take when powered by AI. In 27.5 hours, sixteen developers had produced 4,962 reports across 390 Bitcoin-related projects.

Among them were 85 issues rated critical and 635 high severity. This does not mean that Bitcoin suddenly had 85 critical exploitable flaws. The majority of these reports still needed to be reproduced and verified by humans. An AI can misunderstand code, invent an exploitative condition, or consider perfectly intended behavior vulnerable.

During the first assessment, only part of the thousands of results had been confirmed. The volume remains unusual. Security teams can now run multiple models simultaneously across hundreds of software repositories and obtain thousands of leads in a matter of hours.

The problem then shifts to sorting. You must reproduce the bug, determine the versions affected, measure its impact, prepare a fix and verify that the fix does not create another problem. Bitcoin Core encountered this phenomenon directly in 2026. While preparing for version 32.0, contributor Matthew Zipkin used Kimi K3 to examine the new HTTP server. The model identified a scenario where certain queries could cause excessive memory consumption.

The developers then resumed the work manually. The bug was real, but human analysis showed its scope exceeded the AI’s initial diagnosis. This is probably a good picture of the current situation: the model finds a lead quickly, then the developers have to figure out what it actually means. AI accelerates the first step. It does not delete the following ones.

Lightning and wallets have already received the warning

Bitcoin Core is just one part of the Bitcoin ecosystem. Millions of dollars can depend on a wallet, a Lightning node, a payment processor or a software library without a breach directly affecting the Bitcoin protocol.

The summer of 2026 provided several examples. Core Lightning received a series of reports generated with the help of AI models. Several vulnerabilities have proven serious enough to prompt developers to ask operators to immediately update their software.

Those who couldn’t install the patch quickly were even asked to cut off communications from their nodes. Technical details were kept under embargo for two weeks. This is exactly the kind of situation that becomes trickier if another model can regain vulnerability before this period ends.

Coldcard had another experience. After a theft exceeding $114 million linked to a random generation weakness, the manufacturer used multiple AI models to search for other vulnerabilities.

Analyzes identified additional issues in transaction management, USB exchanges and the update mechanism.

Again, AI has not replaced human auditing. It dramatically increased the amount of code that could be examined. This ability obviously works both ways. A company can use a template to secure its wallet. An attacker can send the same GitHub repository to a comparable model and ask it where the errors are. Open source software does not become less secure because of this. But the speed at which its flaws can be explored is increasing.

Bitcoin must now correct before AI rediscovers

For Bitcoin node operators, the change should especially make intermediate releases more important. A minor update may now contain a fix for a vulnerability that is easy enough to find that Bitcoin Core refuses to wait for the next major release.

Users who stay on older versions for a long time take more risks. Bitcoin Core does not maintain all of its versions indefinitely. Once a branch reaches end of life, it generally no longer benefits from new fixes.

The arrival of AI therefore reinforces advice that already existed: keep the Bitcoin ecosystem software up to date. It also changes the way we measure the danger of a bug. For a long time, a vulnerability could remain hidden for several years simply because no human had taken the time to precisely examine the few lines affected.

This reasoning becomes less reassuring. A more efficient model can appear six months later and analyze in a few minutes code that has remained intact for five years.

This absolutely does not mean that AI can “break Bitcoin”. It can’t guess private keys, break SHA-256, or arbitrarily change the 21 million BTC limit. The warnings about AI and Bitcoin today mainly concern software, much more than the cryptographic primitives themselves. The difference is important.

A network can have strong cryptography while still having errors in an HTTP server, wallet, library, or Lightning Network implementation. LLMs are increasingly excelling in examining this type of code. Bitcoin Core therefore adapts its defense to a fairly new situation: a vulnerability can be officially secret while still being technically easy to find.