London: The Kali Linux team has published a new guide enabling security professionals to run AI-assisted penetration testing entirely on local hardware, with no data sent to cloud services. The setup lets users issue penetration testing commands in plain language, with an on-device AI model interpreting those instructions and executing them through a suite of standard security tools all without an internet connection or third-party API subscription.
Privacy and operational security concerns have long made cloud-dependent AI tools a liability in sensitive testing environments. Regulated industries, government contractors, and red teams operating in air-gapped networks routinely cannot route sensitive data through external services.
The new Kali Linux stack directly addresses that gap by combining three open-source tools: Ollama, a local AI model runtime; mcp-kali-server, a bridge already available in Kali’s repositories that connects the AI to the operating system’s security toolset; and 5ire, an open-source AI assistant that ties the two together into a single working interface.
The stack runs on a consumer-grade NVIDIA GPU. This keeps the hardware barrier accessible for individuals and small teams. Once configured, a security professional can describe tasks in plain English. For example, they can request a scan of a target host for open ports.
The AI interprets the request and selects the correct tool. It executes the task and returns structured results. All processing happens locally. The guide validated this setup with a live port scan. The test confirmed full GPU-accelerated and offline operation.
The release follows Kali Linux’s February integration of Claude AI for penetration testing via the Model Context Protocol a cloud-connected setup that this new guide complements for operators who require complete data sovereignty.
The two guides together position Kali Linux as the most AI-forward penetration testing distribution available, giving practitioners a clear choice between cloud-powered intelligence and fully local operation depending on their environment and compliance requirements.
As AI-assisted offensive security tooling matures rapidly with platforms like Armadin and JetStream Security raising hundreds of millions to automate enterprise defense the availability of open-source, privacy-preserving alternatives for individual researchers and smaller teams is becoming increasingly significant.
Previous
