Key Highlights
- An artificial intelligence initiative identified CVE-2026-117348, a denial-of-service vulnerability in Bitcoin Core that allowed attackers to stall nodes for over 12 seconds per manipulated message.
- Discovered by Spiral’s Project Loupe and validated by Gregory Sanders, the flaw was patched in Bitcoin Core 32.1 and backported across supported releases.
- The disclosure follows Bitcoin Core’s October policy change aimed at expediting disclosures for vulnerabilities easily surfaced by large language models.
Bitcoin Core Patches Node-Stalling Denial-of-Service Flaw
Bitcoin Core developers have officially disclosed and patched a denial-of-service vulnerability cataloged as CVE-2026-117348. The flaw stemmed from an unaddressed inefficiency in how the client processes network inventory messages. While standard Bitcoin inventory messages are capable of holding up to 50,000 entries, the client software did not discard duplicate transaction announcements within the same batch. Consequently, a malicious actor could repeatedly broadcast an orphan transaction with thousands of missing parent transactions, forcing the target node to execute identical, computationally demanding validation routines thousands of times in succession.
Exploiting this mechanism enabled an attacker to stall node processing for more than 12 seconds with a single crafted message. During these stalls, the affected node was rendered temporarily incapable of handling newly broadcast blocks or processing routine peer-to-peer network traffic. Despite the disruption to node responsiveness, the official advisory clarified that the issue did not compromise private keys, facilitate the theft of bitcoin, or cause permanent damage to node data.
Discovery and Remediation Timeline
The vulnerability was originally uncovered by Project Loupe, an automated security framework powered by large language models (LLMs) developed by Elias Rohrer at Spiral, the open-source Bitcoin arm of Block. Following the machine-assisted discovery, Bitcoin Core developer Gregory Sanders verified the technical finding and escalated it to the project’s security team on July 8.
On August 4, an independent security researcher using the handle Stutxo filed a separate report regarding the same behavior. That same day, Sanders submitted a corrective pull request to the Bitcoin Core repository. Core maintainers merged the fix on August 6, subsequently rolling it into the general release of Bitcoin Core 32.1 on October 27. The formal public disclosure arrived on November 20 alongside patch releases for maintained legacy branches. The deployed fix resolves the issue by ensuring each distinct transaction announcement is evaluated only once per inventory message.
The Rise of AI-Assisted Security Audits Across Web3
The identification of CVE-2026-117348 marks another milestone in a transformative period for artificial intelligence in blockchain security research. Throughout the year, automated tools and LLM-driven scanners have surfaced long-standing bugs across several high-profile cryptocurrency networks:
- XRP Ledger: In February, an AI-supported security evaluation detected a critical signature-verification flaw within the proposed Batch amendment prior to its mainnet activation.
- Ethereum Ecosystem: Researchers uncovered a high-severity flaw within the Nethermind execution client that posed operational risks to Ethereum validators.
- Zcash: In late May, researcher Taylor Hornby leveraged an AI-assisted auditing framework to pinpoint a four-year-old vulnerability in Zcash’s Orchard privacy protocol that could have allowed the creation of undetectable counterfeit ZEC before developers quietly remediated it.
- Multi-Project Sprints: During an August initiative, 16 volunteer developers deployed automated scanning across 390 Bitcoin ecosystem repositories, yielding 4,962 initial findings—including 85 flagged as critical—in roughly 24 hours.
While automated tools have accelerated the discovery of critical bugs, they have also placed significant operational strain on open-source maintainers. Engineering teams, including developers supporting Core Lightning, have had to manage substantial influxes of AI-generated bug reports, filtering out noise and false positives to isolate actionable vulnerabilities requiring code changes.
Why This Matters
The speed with which LLMs can parse legacy codebases has disrupted conventional assumptions about software vulnerability lifecycles. Security analysts warn that the same automated tools aiding defensive auditors are accessible to malicious actors seeking zero-day exploits in mature code. This dynamic significantly reduces the time window between vulnerability existence and potential exploitation.
In response to this shifting threat landscape, Bitcoin Core updated its vulnerability disclosure framework in October. Under the revised guidelines, vulnerabilities categorized as easily discoverable—specifically those detectable through relatively simple LLM-assisted analysis—are prioritized for accelerated public disclosure alongside corrective software builds. CVE-2026-117348 represents the operational application of this accelerated strategy, emphasizing rapid patch deployment to minimize exposure windows across the node ecosystem.
Frequently Asked Questions
Did CVE-2026-117348 put user funds or Bitcoin wallets at risk?
No. The vulnerability was strictly a localized denial-of-service vector that temporarily impaired a node’s ability to process new blocks and incoming communications. It did not create any mechanism for unauthorized transactions, fund theft, or permanent state corruption.
How does the patch fix the vulnerability in Bitcoin Core?
The corrective update, merged ahead of Bitcoin Core 32.1, alters how incoming inventory batches are parsed. The node now enforces deduplication, ensuring that any single transaction announcement is evaluated only once per inventory message, neutralizing the repetitive compute loops used to freeze nodes.
Why did Bitcoin Core adjust its vulnerability disclosure policy?
In October, Bitcoin Core revised its policy to address the increasing prevalence of LLM-driven vulnerability research. Because AI tools can rapidly uncover latent bugs, the project now accelerates the disclosure and release schedule for flaws that are easily detectable, reducing the likelihood that adversaries discover and weaponize them first.




