Bitcoin Red Team using Chinese AI models including Moonshot AI's Kimi K3 to identify bugs in Bitcoin's open-source codebase.
Tech & Launches ·
The Bitcoin Red Team, led by pseudonymous developer Calle, is deploying Chinese AI models including Moonshot AI's Kimi K3 to systematically scan Bitcoin's open-source ecosystem for security vulnerabilities. According to reporting, the volunteer effort combines automated analysis with human review across wallets, Lightning applications, libraries, and related projects. In August, the group identified 4,962 findings across 390 projects, with 85 classified as critical severity and 635 as high severity; developers have confirmed numerous real critical and high-level flaws, though affected projects remain unnamed.
The Red Team adopted Chinese models after encountering restrictions with American counterparts. OpenAI and Anthropic impose limitations during security research that Calle described as constraining, prompting the shift to locally-runnable alternatives like Kimi K3, which can process large codebases with minimal human direction. The team has also utilized GLM 5.2 from Chinese developer Z.ai. Lightning applications proved particularly complex to audit and were described as exhibiting more vulnerabilities than average.
Response times to reported findings vary significantly across projects, Calle noted, reflecting underlying health differences. Projects that began AI audits months prior occupy "a completely different position" than those without such efforts. Calle warned that unmaintained projects warrant caution and urged teams to establish ongoing AI audit pipelines. Despite characterizing the current state as dire, Calle framed the audits as ultimately strengthening Bitcoin's codebase.