Media
1 item
Description
Checks AI answers for fabricated references, leaked data, and unverified claims before you trust them. Hardin Filter checks AI-generated answers for reliability before you trust them. Highlight any AI-generated text, right-click, and select "Hardin Filter: Check Selected Text." In seconds you get a clear report showing whether the answer is reliable, questionable, or contains outright errors — plus a signed, independently verifiable receipt. WHAT IT CHECKS: • Fabricated references — URLs, packages, and papers that don't actually exist • Leaked data — secrets, API keys, PII, or system prompt leaks in the answer • Source attribution — whether claims are backed by reachable sources • Overconfident tone — sounds certain but has no supporting evidence • Numeric consistency — numbers that don't add up • Internal contradictions — statements that conflict with each other • Recency — time-sensitive claims that may be outdated • Known-entity accuracy — catches AI confidently misstating real products/protocols HOW TO USE: 1. Go to any AI chat (ChatGPT, Copilot, Claude, Gemini, Kimi, or any other) 2. Select/highlight part of the AI's answer 3. Right-click → "Hardin Filter: Check Selected Text" 4. Read the result: Checked (green), Review Advised (amber), or Issues Found (red) On supported sites (ChatGPT, Claude, Gemini, Copilot), the filter also runs automatically and badges each answer. Every check produces a cryptographically signed receipt — tamper-proof, independently verifiable evidence of what was checked and what was found. PRIVACY: • Selected text is sent to the Hardin Filter API for checking • No personal data is stored — text is hashed, not retained in plain form • No tracking, no ads, no data selling Built on TBN Protocol — open source trust infrastructure for AI agents. Learn more: https://hardinfilter.com