Media
4 items
Description
Verify selected AI knowledge claims with Spark KNE. Spark KNE Verify helps users inspect selected AI and web knowledge claims across daily tools. The extension sends selected text to Spark KNE only after explicit user action, then returns a cautious trust state based on source evidence and KNE Trust Policy. Version 0.1.8 adds clearer current-event and mixed-result explanations, keeps background candidates separate from supporting evidence, and records a consented result-viewed event so Spark can measure whether users actually reach the verification result. Interaction telemetry contains only event metadata such as entry point, a text-length bucket, request identifier, trust state, and verification context; selected text and page URLs are sent only when the user explicitly verifies. Anonymous trial verification remains limited to 10 times per 5 hours before login; feedback/history/contribution remain behind login. Current trust states include: - Source supported - Evidence insufficient - Review needed - Conflicted - Outdated - High risk - Not knowledge Users can also submit source hints, challenges, and corrections to the KNE review queue. User submissions do not directly change KNE Core data; they require review. Current test scope - Active selected-text verification is the primary workflow. - Users can right-click selected text on AI pages and ordinary web pages; floating selection UI may appear on supported pages. - Public test registration is capped at 200 ordinary public users. Privacy summary - Privacy policy URL: https://knespark.com/privacy.html - Spark verifies only text the user actively selects or sends through the extension. - Spark collects account email, account ID, authentication state, anonymous trial identifier, usage quota, selected verification text, source page URL, page title when available, platform label, contribution submissions, feedback submissions, blocking issue reports, consented interaction events, timestamps, trust-state results, review status, quota counters, diagnostic events, IP address, user-agent, and operational logs needed for the beta service. - Spark processes selected text to classify knowledge claims, compare source evidence, detect gaps or conflicts, return cautious trust states, enforce anonymous trial and account usage limits, debug errors, prevent abuse, review user contributions, review feedback, and resolve reported blocking issues. - Spark may store account records, local extension settings, anonymous trial identifier, authentication token, consent state, and up to 30 consented interaction events awaiting retry in Chrome storage, plus anonymous trial usage, Verify history, contribution records, feedback, blocking issue reports, usage counters, review status, analytics events, source gap records, and operational logs on Spark servers. The retry queue does not contain selected verification text. - Spark does not sell, rent, trade, or share user data with advertisers or data brokers. Data is sent only to Spark-controlled services at https://knespark.com and infrastructure processors, including Vultr cloud servers and the user-controlled browser/Chrome extension storage environment, needed to host, secure, monitor, back up, and operate the service. - Spark does not collect GPS, precise location, or address data. Server and security infrastructure may process IP address as operational log data for security, reliability, fraud prevention, abuse prevention, and troubleshooting. - Spark does not monitor clipboard content. - Spark does not run background conversation recording. - Spark stores anonymous trial identifiers and authentication tokens locally in Chrome storage. - Consented interaction events identify actions such as extension open, selection captured, Verify clicked, verification result viewed, and first Verify outcome. Event metadata may contain the entry point, selection presence, a coarse length bucket, request identifier, trust state, and verification context; it does not contain selected text or page URL. - Submitted verification text, source page information, platform labels, contribution content, feedback, and blocking issue reports are sent to https://knespark.com only after the corresponding explicit user action for KNE verification, behavioral beta cohort analysis, review, beta operations, and product improvement.