On Nov. 20, 2026, Bitcoin Core announced a medium‑severity security issue that lets a hostile peer temporarily halt a node’s operation for more than a dozen seconds per malicious inventory packet.

BTC / USD
Source: CoinGecko. Prices are indicative.

Technical description and impacted releases

The vulnerability, listed as CVE‑2026‑117348 and named “Node stall through duplicate orphan announcements,” stems from the client’s handling of the orphanage, a temporary store for transactions lacking parent data.

When an inventory message contains up to 50,000 entries, the software previously failed to ignore repeated announcements of the same orphan transaction, forcing the node to repeat expensive dependency checks for each duplicate.

An attacker can repeatedly broadcast an orphan transaction with many missing parents, causing the node to spend over 12 seconds on each duplicated entry and preventing it from processing new blocks or transactions during that period.

The bug affects Bitcoin Core releases 29.0‑29.5, 30.0‑30.4 and 31.0‑31.2; the legacy 29.x line will not receive further updates.

Patch versions 30.5, 31.3 and 32.1 address the problem, with version 32.1 delivering the fix on Oct. 27 and the Nov. 20 advisory accompanying updates for the still‑supported older branches.

AI‑assisted discovery and remediation workflow

The flaw was first identified by Project Loupe, an automated large‑language‑model security platform created by Elias Rohrer at Spiral, the open‑source arm of Block.

Bitcoin Core maintainer Gregory Sanders confirmed the finding and opened a corrective pull request on Aug. 4; an independent researcher known as Stutxo filed a parallel report the same day.

Developers merged the correction on Aug. 6, and the patched code was incorporated into the October 27 release of Bitcoin Core 32.1.

This sequence demonstrates how AI‑driven scanning can surface deep‑seated bugs that have eluded conventional code reviews for years.

Wider AI‑driven security activity in 2026

The node‑stall case follows several AI‑related disclosures across major blockchain platforms earlier this year.

In February, an AI audit revealed a critical signature‑verification weakness in a pending XRP Ledger amendment.

Later, an AI review uncovered a high‑severity issue in Nethermind, an Ethereum execution client, that could have affected validator performance.

By May, researcher Taylor Hornby used an AI‑based tool to locate a four‑year‑old flaw in Zcash’s Orchard protocol that might have allowed undetectable counterfeit coins; developers patched it before any public release.

In August, a coordinated Bitcoin security sprint generated 4,962 findings across 390 projects in a single day, with 85 initially flagged as critical, highlighting the volume of alerts AI can produce.

In response, Bitcoin Core updated its vulnerability disclosure policy in October, opting for faster public announcements for bugs that are readily discoverable with LLM assistance, a policy applied to CVE‑2026‑117348.

Security implications for open‑source blockchain software

The advisory clarifies that the bug enables only temporary denial‑of‑service attacks and does not allow theft of funds or permanent node damage.

Security experts warn that the same AI techniques used by researchers could be repurposed by malicious actors to locate exploitable weaknesses in older codebases, potentially shortening the useful life of traditional audit cycles.

The episode illustrates an emerging security paradox: AI can accelerate the detection of overlooked vulnerabilities, yet its accessibility also equips attackers with comparable discovery capabilities.