Bitcoin Red Team uses Chinese AI models to scan the open-source stack for flaws

The Bitcoin Red Team, a volunteer security group, is using Chinese AI models alongside human review to search for vulnerabilities across nearly the entire Bitcoin open-source ecosystem, according to pseudonymous developer and team lead Calle. The work covers wallets, Lightning applications, software libraries, and other Bitcoin projects, with credible findings privately disclosed so developers can patch issues before details are published. Calle said the team has used Moonshot AI’s Kimi K3, Z.ai’s GLM 5.2, and models from OpenAI and Anthropic. He argued that U.S. models often impose restrictions that get in the way of security research, prompting the team to lean more heavily on Chinese alternatives that can be run locally. The group said it has completed a basic scan of virtually all Bitcoin open-source code and, in August alone, filed 4,962 findings across 390 projects. Of those, 85 were rated critical and 635 high severity. Calle said developers had confirmed many real critical and high-risk issues, though the team has not identified affected projects or released technical details. He also said Lightning software has been especially hard to review because of its complexity, and warned that projects without their own AI audit pipeline may be falling behind.