The opt-in service uses the company’s most advanced models to find flaws in open source projects and has already identified thousands of vulnerabilities in widely used software.

Anthropic announced OSS Scanner, a free service that periodically scans open source repositories for security flaws using its most advanced models, including Claude Mythos.

According to the company itself, each report includes a proof of concept of the exploit, an explanation of the flaw, and, when possible, a suggested fix. Anthropic described the launch as the result of the experience accumulated during Project Glasswing, a previous initiative using Claude to hunt for vulnerabilities.

The process begins by setting up an isolated environment for each project participating in the program. A Dockerfile installs all dependencies and compiles the code, allowing the AI agent to conduct the audit without internet access.

#Tecnologia #InteligenciaArtificial #Regulation