Bitcoin Red Team Uses AI to Find Dozens of Vulnerabilities Across Core Projects

A volunteer security initiative known as the Bitcoin Red Team has used frontier AI models to scan 150 Bitcoin repositories, identifying more than a dozen vulnerabilities in various projects. The team, led by AnchorWatch CEO Rob Hamilton, has spent around $20,000 on AI services, including Kimi K3, OpenAI's GPT Sol, Anthropic's Claude Fable, and Opus models, as well as Z.ai's GLM 5.2.

The group's efforts have been successful in finding critical exploits, with Calle, a pseudonymous Bitcoin developer, stating that they are averaging one critical exploit per hour per person. The team has reported vulnerabilities to several projects within the last 12 hours.

Hamilton noted that funding is secured for the initiative and that it is not necessary to donate money. He also mentioned that the group has been connected with OpenAI for assistance in managing the Cyber Harness, a more expensive scan but one that yields good results for load-bearing portions of the Bitcoin ecosystem.

The use of AI in finding security flaws across the crypto industry continues to grow. Earlier this year, researchers using Anthropic's Claude Opus 4.8 uncovered a four-year-old flaw in Zcash that could have allowed attackers to create unlimited counterfeit ZEC. In August, Coinkite stated that it believes attackers used AI to identify the Coldcard wallet vulnerability.