← Editorial Bitcoin

Bitcoin Red Team Uses AI to Identify Over a Dozen Critical Vulnerabilities

A volunteer security initiative says it deployed AI models to scan 150 Bitcoin repositories and uncovered more than a dozen vulnerabilities, while building an open-source platform for automated security audits.

· +1

A Bitcoin-focused red team, led by AnchorWatch CEO Rob Hamilton, has deployed multiple AI models to scan 150 Bitcoin repositories, according to an announcement shared this week. The group reported identifying more than a dozen vulnerabilities and has spent approximately $20,000 on AI services to date.

The initiative uses frontier AI models including Kimi K3, OpenAI's GPT Sol, Anthropic's Claude Fable and Opus, and Z.ai's GLM 5.2 to probe for security flaws and generate documentation. The team also integrated OpenAI's Cyber Harness for scanning critical infrastructure components, though at higher cost.

Pseudonymous Bitcoin developer Calle reported that the red team has developed multiple AI-powered review systems targeting wallets, cryptographic libraries, and other Bitcoin projects. According to Calle, the team is discovering critical exploits at a rate of approximately one per hour per person and reports burning through $10,000 daily in scanning costs. The team has reportedly disclosed critical vulnerabilities to several projects within a 12-hour period but has not yet detailed which projects were affected.

The development reflects a broader trend of AI being used to identify security flaws across crypto infrastructure. Earlier this year, researchers uncovered a four-year-old flaw in Zcash using Claude Opus. In August, Coinkite suggested attackers used AI to identify the Coldcard wallet vulnerability, while the Boltz Bitcoin bridge suspended operations after reporting that attackers were using AI to find vulnerabilities faster than patches could be deployed.

How this piece reads Neutral tone +1
Site call on Bitcoin Buy score 37.2

These two are not the same thing, and one does not produce the other. The left is how this single article reads, from its tone alone. The right is the site’s own call on the asset, from indicators and analysis. Press tone feeds no score and no signal: on the only corpus this site has measured, daily tone tracked the move that had already happened and showed no measurable link with what followed.

More on Bitcoin All pieces →

Rewritten from the headline, the teaser and the one-line summary the qualification step produced — that is all the material there is, and nothing is added to it. The source link is kept on file so any item can be checked, and is not published here.