Bitcoin Core Fixes AI-Discovered Flaw That Could Stall Node Processing
Bitcoin Core developers disclosed a vulnerability uncovered through AI-based security research and released patched versions of the software. The flaw could temporarily stall a node’s transaction and block processing if an attacker sent crafted messages.Bitcoin.com reported on November 22 that Bitcoin Core developers disclosed the medium-severity vulnerability, CVE-2026-117348, on November 20. The issue works by flooding a node with duplicate transaction announcements, forcing it to repeat unnecessary computation. A single message can delay processing by more than 12 seconds.The bug does not allow attackers to steal Bitcoin or cause permanent damage to nodes. Affected versions are 29.0 through 29.5, 30.0 through 30.4, and 31.0 through 31.2. The fix was included in versions 30.5, 31.3 and 32.1. No patch will be issued for the unsupported 29.x branch.The vulnerability was first identified by Project Loop, a security research project developed by Elias Loerer of Spiral, Block’s open-source Bitcoin development organization. The project uses large language models, or LLMs, to find security flaws. Bitcoin Core developer Gregory Sanders verified the finding and reported it to the security team.The problem originated in the orphan pool, which temporarily stores transactions whose prerequisite transactions have not yet been received. Because the software did not adequately filter duplicate announcements, it could repeat the same verification work. Attackers could increase the processing burden on a node by exploiting the fact that a single message can contain as many as 50,000 entries.As AI speeds up vulnerability discovery, disclosure practices are also changing. Bitcoin Core revised its disclosure policy in October to quickly announce flaws that can be found relatively easily through LLM analysis once patched versions are released. The move reflects the risk that attackers can use the same tools as defenders to identify vulnerabilities.