AI in Bitcoin Security
In a recent announcement, a volunteer group dedicated to security has revealed that it employed advanced artificial intelligence (AI) models to examine 150 Bitcoin repositories. This endeavor uncovered over a dozen vulnerabilities as developers increasingly turn to AI for blockchain auditing.
Investment and Team Formation
Rob Hamilton, CEO of AnchorWatch, shared insights on the project via X, highlighting that the initiative has invested around $20,000 in AI technology while establishing a “Bitcoin red team” task force.
“We have been working tirelessly, with spending of approximately $20,000 spread across various services so far,” Hamilton stated.
He assured that funding is secured and expressed gratitude for offers of donations, noting they are not needed. The red team, comprised of cybersecurity experts, is instrumental in simulating malicious attacks on software to identify weaknesses before they can be exploited.
Tools and Collaboration
Hamilton detailed that the Bitcoin red team employs tools including Kimi K3, OpenAI’s GPT Sol, Anthropic’s Claude Fable, Opus models, and Z.ai’s GLM 5.2 to pinpoint vulnerabilities and create necessary documentation. He mentioned a collaboration with OpenAI to manage the Cyber Harness project, which entails a more costly scanning process but is deemed essential for critical facets of the Bitcoin ecosystem, reporting promising outcomes thus far.
Community Insights
Calle, a pseudonymous developer in the Bitcoin community, remarked that the initiative has developed numerous AI-based review systems aimed at wallets, cryptographic libraries, and various Bitcoin projects.
“On average, we’re detecting one critical vulnerability each hour for each team member,” Calle informed followers on X. “Just within the last twelve hours, we’ve reported significant vulnerabilities to multiple projects. Fortunately, this process is expensive and we are incurring around $10,000 in costs daily.”
Broader Trends in AI and Cryptocurrency Security
While the team has not disclosed the specific projects affected or the details of the vulnerabilities found, this announcement aligns with a broader trend of increasing AI involvement in identifying security issues throughout the cryptocurrency sector. Earlier this year, for instance, researchers utilizing Anthropic’s Claude Opus 4.8 identified a significant four-year-old flaw in Zcash, which could have enabled the creation of unlimited counterfeit ZEC.
Furthermore, in August, Coinkite indicated a belief that attackers were leveraging AI to discover vulnerabilities in the Coldcard wallet, while Bitcoin bridge Boltz paused its swap services after reporting that AI was being used to exploit weaknesses more rapidly than they could be repaired.
