Bitcoin’s AI security sprint found 6,700 issues in 55 hours, but no one knows how many are real
Bitcoin Red Team said its AI-assisted security sprint produced 6,700 findings across 425 Bitcoin-related projects in its first 55 hours, with 1,029 labeled high or critical. Earlier at 27.5 hours, it had 4,962 findings across 390 projects. The campaign showed how AI can rapidly scan an ecosystem, but it has not published audit-ready denominators, false-positive rates, fix rates, or case-level outcomes, so the share of alerts that became confirmed vulnerabilities is unknown. Rob Hamilton said models did the broad analysis while experts handled prompting, reproduction, disclosure, and triage, and that small amounts of subject-matter context could change severity judgments. Reported bottlenecks were operations, disclosure handoff, and triage. The campaign also said only 19.5% of scanned projects had SECURITY.md files and 13.1% listed an email, but the measurement method is unclear. Overall, the sprint demonstrates scale and speed, not confirmed security impact.
