Bitcoin Red Team uses Chinese AI to hunt vulnerabilities in software ecosystem

A volunteer collective of security specialists and developers has formed the Bitcoin Red Team to proactively scan the network's software ecosystem for vulnerabilities discoverable by artificial intelligence. The group aims to identify and patch flaws in the applications and services surrounding the protocol before they can be weaponized by malicious actors.

Consisting of approximately 20 to 25 members, including developers such as Calle and James O'Beirne, the group focuses specifically on the software ecosystem—including wallets and third-party services—rather than the core Bitcoin protocol, which the team maintains remains secure. The initiative was spurred by recent security incidents, most notably a firmware and random number generator (RNG) vulnerability discovered in Coldcard air-gapped wallets.

The AI Arms Race

The group's strategy involves utilizing powerful AI models to simulate attacks. Notably, the team prefers Chinese AI models, such as Moonshot AI's Kimi K3 and GLM 5.2, over American alternatives. According to the group, U.S.-based models frequently employ strict guardrails that block requests related to cybersecurity research, whereas Chinese models provide the necessary flexibility for deep security analysis.

This shift is part of a broader trend where "frontier" AI models are altering the cybersecurity landscape. While AI provides defenders with new tools for bug hunting, it simultaneously eliminates "security through obscurity" by allowing attackers to analyze complex code and generate exploits with unprecedented speed. Because Bitcoin represents a direct financial incentive, it has become a primary target for these automated techniques.

The End of Information Asymmetry

The emergence of these tools represents a paradigm shift in how software is secured. Historically, a significant gap in expertise existed between elite hackers and novice attackers. AI is rapidly closing this gap, allowing individuals without advanced security backgrounds to execute sophisticated exploits.

"Simple exploits can now be completed end to end by someone who doesn't know how to do it without AI," said Calle, a member of the Bitcoin Red Team. He further noted that the previous reliance on information asymmetry to create "security theater" is no longer viable, as AI removes the secrets previously hidden within software code.

Future Outlook

As AI continues to automate the discovery of vulnerabilities in financial software, the Bitcoin Red Team warns that the speed of patching must increase proportionally to prevent systemic losses. The group is currently working directly with open-source projects to remediate the flaws they uncover.

Industry observers will be watching whether this proactive, volunteer-led model becomes the standard for other cryptocurrency ecosystems. The primary challenge remains the constant evolution of AI capabilities, which may soon allow for the discovery of vulnerabilities that human developers—and current AI defenders—cannot yet conceive.