Bitcoin Core maintainers have updated the project's vulnerability disclosure rules in response to a sharp rise in security reports that artificial intelligence systems can quickly reproduce. The change prioritizes rapid patching and limited public details for certain classes of issues rather than prolonged secrecy.

Antoine Poinsot, a developer affiliated with Chaincode Labs, announced the revision on the Bitcoin Development Mailing List. He noted a recent increase in reports involving flaws that are straightforward to rediscover, largely because large language models have lowered the barrier to identifying specific types of problems or spotting covert fixes.

Shift from Long Embargoes

Previously, Bitcoin Core balanced transparency against the need to give users time to upgrade. The project's slow upgrade culture led maintainers to favor extended embargo periods. Security fixes generally could not be backported and became available only with the next major release.

For vulnerabilities classified as easy to discover, that approach is no longer suitable. During triage, maintainers may flag an issue as readily findable. In practice, this covers problems that a simple prompt to a large language model can uncover, though the category is not limited to AI-assisted finds.

Such issues will now receive a fix backported to every currently maintained release. A point release will follow as soon as possible, accompanied by a security advisory that includes some details. The updated policy is published on the Bitcoin Core website.

Disclosure Timelines for Other Severities

Low-severity vulnerabilities will be disclosed two weeks after a major version containing the fix is released. Medium- and high-severity issues will be disclosed two weeks after the last affected software version reaches end of life, typically about one year after the fix appears in a major release.

Critical bugs fall outside the standard schedule and would require case-specific procedures. Poinsot acknowledged that the faster path for easily discoverable issues means the fixes may receive less extended testing before reaching users.

Not all contributors fully agreed on the details. One longtime security researcher preferred keeping the option to ship a fix without an advisory and withhold details on a case-by-case basis. Poinsot rejected that approach, arguing that an alert without information could prompt unnecessary rushed upgrades by unaffected users.

Context of Rising AI-Assisted Reports

The policy revision follows a summer marked by heightened AI-driven scrutiny of Bitcoin-related software. A volunteer group known as the Bitcoin Red Team applied large language models to 390 projects and recorded nearly 5,000 potential findings, including 85 rated critical. Only a portion of those reports had been independently verified at the time.

In the same period, Core Lightning developers advised node operators to upgrade or temporarily disconnect after receiving multiple AI-generated vulnerability reports that proved valid. Technical details were held back for two weeks while updates were prepared. Separately, a privately reported flaw in BTCPay Server was exploited on live merchant systems, draining Lightning nodes used for payments.

A separate incident involving Coldcard hardware-wallet firmware, unrelated to the AI reports, resulted in the theft of hundreds of bitcoin valued at roughly $114 million across multiple sweeps. The manufacturer later noted that AI tools had assisted in identifying additional issues during the remediation process.

Luke de Wolf, host of a Bitcoin security podcast and author of a book on defending the network, described the Core update as a positive step even though he had disagreed with Poinsot on security approaches in the past.

No vulnerability has yet been handled under the new classification. Maintainers indicated the adjustment aims to reduce the window of exposure for users while adapting to tools that can regenerate certain secrets almost instantly.