AI Red Team Finds Hundreds of Bitcoin Exploits

A volunteer security initiative operating as a Bitcoin red team identified hundreds of critical exploits after scanning 150 Bitcoin repositories using frontier AI models. Led by Rob Hamilton, the team utilized models from OpenAI, Anthropic, Z.ai, and Kimi K3 to probe cryptographic libraries, wallets, and infrastructure.

In the first 29.8 hours of operation, the group discovered 4,962 potential issues across 390 projects. Of these, 720 were classified as high- or critical-level vulnerabilities, with 21.4% of the findings proven reproducible. The team has already notified several projects regarding these critical flaws.

The operation is resource-intensive, with reported spending of approximately $10,000 per day on AI services. This effort follows a broader trend of using artificial intelligence to uncover vulnerabilities in the cryptocurrency industry, similar to previous flaws found in Coldcard and Zcash wallets.