HateBlocker

HateBlocker quietly hides toxic comments on YouTube and Facebook before you read them. It works entirely on your device. Chrome's built-in Gemini Nano language model classifies each comment as toxic or not, and HateBlocker hides the toxic ones with a small "click to reveal" placeholder. No comment text, no browsing data, no analytics are ever sent to a server. ✓ Filter YouTube and Facebook comments ✓ 100% on-device — uses Chrome's built-in AI (Prompt API) ✓ No accounts, no API keys, no subscriptions ✓ Three sensitivity levels: Strict, Balanced, Lenient ✓ Choose which categories to filter: personal attacks, hate speech, threats, sexual harassment, profanity, spam ✓ Three hide styles: a subtle placeholder, a prominent banner, or an outline-only mode that keeps the comment visible with a flag for review ✓ Per-platform toggles ✓ Re-hide a revealed comment with one click ✓ Re-scan the page on demand ✓ Counter of how many comments have been hidden ### Requirements - Chrome 138 or later - A device that meets Chrome's requirements for the on-device AI model (the first time you use HateBlocker the model — about 1.5 GB — will download in the background) - The model status is shown inside the extension popup with a one-click download button when needed ### What HateBlocker is not - It is not a perfect classifier. Gemini Nano is small and English-first. It catches obvious toxicity reliably but will miss sarcasm and subtler attacks, especially in other languages. - It fails open. If the model is unsure or the classification fails, comments stay visible. - It is not a content-moderation tool for platform owners — it just hides comments for you, locally, in your own browser. ### Privacy HateBlocker does not collect, transmit, store, or share any browsing data. All processing is local. See the privacy policy for details.