Nov 21, 2026, 1:16 a.m. EST
2 min read
Summary
- An artificial intelligence application created within Jack Dorsey’s company Block uncovered a Bitcoin Core vulnerability capable of briefly halting a node from handling network messages by duplicating a transaction identifier 50,000 times.
- Malicious actors could extend the interruption by sending more messages, which might delay exchange deposits or force miners to operate using obsolete blockchain data.
- Bitcoin Core classified this security flaw as medium severity and deployed patches in versions 30.5 and 31.3 guaranteeing that every transaction mentioned in a batch is evaluated only a single time.
An automated AI utility developed by Jack Dorsey’s firm Block discovered a method to momentarily freeze a bitcoin node from managing incoming network messages by supplying a single transaction ID copied 50,000 times.
Bitcoin Core, the program utilized to operate these nodes, made the vulnerability public on Friday designated as CVE-2026-117348 while simultaneously launching versions 30.5 and 31.3 containing patches.
The notification attributes the discovery to an automated AI scan executed via Elias Rohrer’s Project Loupe housed at Spiral, which functions as Block’s open-source development division, as reported by CoinDesk.
Nodes function as computers that verify bitcoin transactions and relay transfers and blocks between one another.
The real-world impacts of this defect proved extensive: A trading platform depending on an impacted node could experience delays when crediting deposits, whereas a miner might expend processing energy operating on an outdated perspective of the blockchain ledger.
Core programmers categorized the flaw as “medium severity” and as “easily discoverable.” In accordance with the security policy established on October 22, easily detectable bugs obtain fixes across maintained releases as quickly as possible, accompanied by a public advisory published alongside the software updates, as documented by CoinDesk.
How the trick worked
Bitcoin nodes store orphan transactions, which represent payments utilizing coins from a prior transfer that the node has not yet encountered. The node subsequently searches to locate the missing components.
Beginning with version 29.0, whenever a different node broadcasts one of these orphan transactions, Bitcoin Core compiles an inventory of its absent parent transactions and inspects its databases prior to determining what data to request.
Nodes broadcast transactions in groups containing as many as 50,000 items. The software previously failed to bypass duplicate items.
An adversary could initially transmit an orphan transaction featuring thousands of missing parents, and then duplicate that transaction ID 50,000 times inside a single broadcast. The recipient node would subsequently execute the resource-intensive verifications repeatedly.
“Each such message stalls the node for over a dozen seconds,” according to the advisory. Throughout that timeframe, the system handles zero messages originating from other nodes, inclusive of fresh blocks. Dispatching additional messages could prolong the operational disruption.
The implemented patch ensures that the node evaluates each transaction announced within a batch only a single time.
This guideline illustrates the apprehension among developers that AI applications make specific vulnerabilities inexpensive for attackers to re-identify, thereby diminishing the safety margins previously maintained by keeping patches confidential.
Read More: Bitcoin Core changes security rules as AI makes finding code bugs dangerously easy
Originally published at https://www.coindesk.com/markets/2026/11/21/bitcoin-nodes-could-be-stalled-by-repeating-one-id-50-000-times-ai-found-the-bug.