ClaimShield Nano is a privacy-first fact-checking assistant that lives directly inside Chrome. It combines Chrome’s built-in Gemini models with the Gemini API to analyze highlighted text, extract the main claim, and assign a verdict: OK, NEEDS_REVIEW, or ABSTAIN. When the user selects text and chooses “Verify with ClaimShield”, the extension runs a local reasoning pass using Chrome’s built-in AI, then optionally calls the Gemini API (via Google AI Studio) in Hybrid mode for a second opinion on harder claims. For each claim, ClaimShield Nano automatically searches Wikipedia, pulls 1–3 relevant articles, and shows them as clear, human-readable sources. The system is explicitly honest about limitations. If the claim is about a very recent event, ClaimShield detects that it is likely beyond the model’s training window, clearly flags “⏳ Model outdated / recent event,” and adds safe pseudo-sources plus a live Google News search link so users can check up-to-date reports themselves. This demonstrates how Gemini models + Google AI Studio + Gemini API can improve information quality without sacrificing privacy. Users keep browsing as usual, select any sentence, right-click Verify with ClaimShield, and immediately see a verdict, confidence score, and sources panel. ClaimShield Nano is a focused, working example of multimodal-ready, explainable AI fact-checking — and a foundation for future agents that will also verify claims in images, screenshots, and long documents.
Category tags: