Crypto Developers Turn to Anthropic AI to Patch Software Vulnerabilities

Anthropic launched its OSS Scanner on October 8, shifting toward an automated model that bypasses traditional, time-consuming human reviews. While the company previously managed vulnerability disclosures through manual verification, the sheer scale of detected issues—over 29,000 potential flaws identified in open-source projects over the last six months—necessitated a faster, AI-driven approach. Nethermind, which maintains the software powering Ethereum network operations, and ZEUS, a self-custodial Bitcoin wallet, are among the first to seek enrollment to secure their transaction infrastructure, private key handling, and Lightning Network connections.

This push for AI-assisted defense comes as the crypto industry faces a surge in sophisticated, automated exploits. In August, Bitcoin swap provider Boltz suspended operations after enduring months of AI-assisted attacks that outpaced their internal security response. Anthropic’s new system provides maintainers with reproducible examples of weaknesses and suggested patches, though the company warns that findings may occasionally include false positives. Participation remains subject to individual approval, with Anthropic prioritizing projects based on their systemic importance and the potential impact on infrastructure security. Other industry participants, including the parent company of Kraken, have already begun integrating Anthropic’s models to bolster their own cybersecurity protocols.