- Published on
Red Team Turns to Chinese AI to Find Flaws
- Authors
- Name
- DeepHorus
The Bitcoin Red Team, a volunteer security research group led by pseudonymous developer Calle, has turned to Chinese AI models to hunt for vulnerabilities across nearly the entire Bitcoin open-source ecosystem. The initiative combines artificial intelligence tools with human review to examine wallets, Lightning applications, software libraries, and other critical Bitcoin projects. Researchers privately report credible findings to developers so that flaws can be patched before details are publicly disclosed. According to Calle, the scale of discovery has been staggering, prompting the warning that “everything is broken, Bitcoin is burning.”
The group has been using Kimi K3, an AI model from Chinese startup Moonshot AI that can be downloaded and run locally, allowing it to analyze vast codebases with minimal supervision. In addition, the team has employed Z.ai’s GLM 5.2, as well as models from OpenAI and Anthropic. However, American models have proven problematic for security research due to imposed usage restrictions. Calle complained on X that the team was “red team rugged by OpenAI cyber again,” noting a preference to avoid asking for permission and announcing a switch to Kimi K3.
Despite the challenges, the audit has made significant progress. In August, the Bitcoin Red Team reported filing 4,962 findings across 390 projects, including 85 rated critical and 635 rated high severity. Calle stated that developers have confirmed “a ton of real critical and high vulnerabilities,” although the affected projects and technical details remain undisclosed. The group has essentially completed a basic scan of virtually the entire Bitcoin open-source landscape, with the “low hanging fruit” already identified.
The findings underscore uneven responses from project maintainers. Calle noted that response speed varies widely and serves as a health indicator for each project, advising developers to “act fast.” Lightning software, which enables faster and cheaper Bitcoin payments, was singled out as particularly complex and “more broken than the average.” The developer warned that projects that have not yet integrated AI auditing pipelines are at a severe disadvantage compared to those that embraced the technology months ago.
The Bitcoin Red Team is not alone in relying on Chinese AI for security work. Last month, Hugging Face used GLM 5.2 to investigate a breach after OpenAI models hacked into its systems and U.S. commercial models refused to analyze the attack logs. This trend highlights the growing role of open-source Chinese AI models in cybersecurity, especially when Western models impose policy restrictions.
Calle stressed that unmaintained projects should not be trusted and that AI has intensified the pressure on developers to keep software secure. Although he described the situation as “burning,” he argued that the audits are ultimately strengthening Bitcoin’s codebase. “Bitcoin is the obvious first target, but the rest of the world will follow shortly,” Calle wrote. “Sometimes old things need to burn so new things can grow on healthy soil.”
