An artificial intelligence tool developed by Jack Dorsey’s Block discovered a Bitcoin network vulnerability that could temporarily prevent nodes from processing incoming messages. The flaw allowed an attacker to overwhelm a node by broadcasting the same transaction ID 50,000 times.
Bitcoin Core, the software used to operate Bitcoin nodes, announced the vulnerability on Friday under the tracking number CVE-2026-117348. The development team simultaneously released versions 30.5 and 31.3, introducing patches to eliminate the security weakness.
The security disclosure credited an automated AI-assisted investigation conducted through Project Loupe, an open-source scanning initiative led by Elias Rohrer at Spiral, Block’s Bitcoin development division, according to CoinDesk.
Bitcoin nodes play an essential role in maintaining the network by verifying transactions and sharing payment information and newly mined blocks with other participants.
If exploited, the vulnerability could have disrupted services across the cryptocurrency industry. Exchanges running affected nodes might have encountered delays in confirming customer deposits, while miners could have continued using outdated blockchain information, potentially wasting computing resources.
Bitcoin Core developers assigned the vulnerability a medium-severity rating and described it as easily discoverable. Under the security guidelines introduced on October 22, developers are expected to provide prompt patches for vulnerabilities that can be readily identified across all maintained software versions, alongside a public security announcement.
How Repeated Transaction IDs Could Freeze Bitcoin Nodes
Bitcoin nodes maintain records of orphan transactions, which are transactions that reference coins from an earlier transaction that the node has not yet received. The software must locate these missing parent transactions before it can properly process the orphaned payments.
Starting with Bitcoin Core version 29.0, the software generates a list of missing parent transactions whenever another node announces an orphan. It then checks its existing records to determine which transactions need to be requested.
However, Bitcoin’s transaction announcement system supports batches containing up to 50,000 entries. The vulnerability emerged because Bitcoin Core did not filter out duplicate transaction IDs before performing these checks.
An attacker could take advantage of this behavior by sending an orphan transaction with thousands of missing parent transactions and subsequently announcing the same transaction ID 50,000 times in a single message.
As a result, the receiving node would repeatedly execute expensive verification checks for identical information, consuming resources unnecessarily.
The Bitcoin Core security advisory explained that each malicious announcement could prevent a node from processing other network messages for more than 12 seconds. This included important updates such as newly announced blocks. Attackers could also send additional messages to extend the disruption.
The newly released software versions resolve the issue by ensuring that every transaction ID in an announcement batch is handled only once, even when the same entry appears repeatedly.
The discovery also demonstrates how AI-powered security tools are changing vulnerability research. Automated scanners can identify weaknesses in widely used software more efficiently, but the same capabilities could allow malicious actors to discover exploitable flaws more easily.
Bitcoin developers’ decision to distribute fixes promptly reflects growing concerns that delaying public patches could leave network participants exposed as AI tools make vulnerability discovery increasingly accessible.
































