Show HN: Cybersecurity Phishing Guard for Chrome using local LLMs for privacy (github.com)

A new Chrome extension called PhishGuard aims to enhance online safety by utilizing locally-run large language models (LLMs) to detect phishing and dangerous web pages without sending any data externally. By integrating with local servers from Ollama and LM Studio, the extension can analyze a webpage's URL, visible text, form inputs, metadata, and even screenshots to provide a risk assessment categorized as Safe, Suspicious, or Dangerous, accompanied by identified red flags. Users can actively scan specific pages or opt for passive monitoring, where the extension automatically evaluates each visited page and displays warnings when potential threats are detected. This development is significant for the AI/ML community as it emphasizes privacy-focused AI applications, allowing users to leverage powerful multimodal models, such as google/gemma-4-26b-a4b, effectively on their local machines. By ensuring that sensitive browsing data is not shared with external servers, PhishGuard addresses growing concerns about online privacy while promoting the use of LLMs in cybersecurity. The combination of detailed risk analyses and real-time protection against phishing illustrates the potential of AI-driven tools in safeguarding users' online experiences.