AI
AI Scanner Finds Bitcoin Core Bug That Could Stall Nodes
Block's Spiral unit deployed an automated AI tool that discovered a vulnerability allowing attackers to freeze node processing by repeating a transaction ID 50,000 times.
Key takeaways
- Block's Spiral unit used an AI tool called Project Loupe to discover a Bitcoin Core vulnerability that could freeze node operations
- The bug allowed attackers to stall nodes for over a dozen seconds by repeating a transaction ID 50,000 times in announcement batches
- Bitcoin Core released patches in versions 30.5 and 31.3 with immediate public disclosure, reflecting new security policies for easily discoverable flaws
- The vulnerability could delay exchange deposits and cause miners to work on outdated blockchain views
- AI-assisted security scanning is reshaping Bitcoin Core's approach to vulnerability disclosure and patch management
AI-Powered Security Scanning Uncovers Node Vulnerability
An artificial intelligence tool developed at Block's Spiral open-source unit has identified a vulnerability in Bitcoin Core that could temporarily prevent nodes from processing network messages. The flaw allowed an attacker to freeze a node's operations by sending a single transaction identifier repeated 50,000 times.
Bitcoin Core disclosed the bug on Friday as CVE-2026-117348 and released patched versions 30.5 and 31.3. The vulnerability was discovered through Project Loupe, an automated AI scanning system led by Elias Rohrer at Spiral, according to details first reported by CoinDesk.
How the Exploit Worked
The vulnerability exploited Bitcoin Core's handling of orphan transactions—payments that reference earlier transactions a node hasn't yet received. When nodes announce transactions, they can send batches containing up to 50,000 entries.
Starting with Bitcoin Core version 29.0, the software builds a list of missing parent transactions and checks its records before requesting data. However, the code failed to skip duplicate entries within a single announcement batch.
An attacker could send an orphan transaction with thousands of missing parents, then announce that same transaction ID 50,000 times in one message. The receiving node would perform expensive validation checks repeatedly for each duplicate entry. According to the advisory, each malicious message could stall a node for more than a dozen seconds, during which it would process no messages from other nodes, including new blocks.
Real-World Impact
Core developers rated the bug as medium severity and easily discoverable. The practical consequences extended across the Bitcoin network: exchanges relying on affected nodes could experience delays in crediting deposits, while miners might waste computational resources working on an outdated view of the blockchain.
The fix ensures nodes process each transaction announced in a batch only once, eliminating the redundant validation checks.
Why It Matters
This discovery highlights both the promise and urgency of AI-assisted security auditing in cryptocurrency infrastructure. Bitcoin Core's security policy, adopted in October, now treats easily discoverable flaws as requiring immediate public disclosure alongside fixes—a shift driven by concerns that AI tools make vulnerabilities cheaper for attackers to rediscover. The traditional approach of keeping fixes quiet until widespread deployment offers less protection when automated systems can scan codebases at scale. For enterprises building on Bitcoin infrastructure, this incident underscores the importance of rapid patch deployment and the growing role of AI in both offensive and defensive security postures.
The vulnerability details were first reported by CoinDesk.