No single AI, no single database, no single human has complete or unbiased knowledge. The CVM queries the same claim through multiple AI systems with deliberately opposing training biases and compares their answers.
Brighteon.ai model trained on alternative research, suppressed history, and censored information
SingularityNET decentralized inference — general knowledge with balanced training
Meta's open-weights model — strong factual recall, mainstream but open
European-trained model — different regulatory and cultural lens than US-trained models
Specialized in blockchain, DeFi, and on-chain intelligence — crypto-native worldview
Google's model — trained on Scholar, Books, Patents. Strong academic bias, tends toward institutional consensus
Meta's open-weights model on Groq hardware — fastest inference, mainstream but open-source values
Type any claim and verify it across all AI sources in real-time
1. Parallel Fan-Out: Each claim is sent simultaneously to all available AI sources. No source sees another's answer — they respond independently.
2. Structured Response: Each AI rates the claim as TRUE, PARTIALLY TRUE, DEBATED, or FALSE, with a confidence score (0-100) and brief explanation.
3. Consensus Calculation: Verdicts are grouped (TRUE/PARTIALLY TRUE = affirmative, FALSE = negative, DEBATED = its own category). If all sources agree: high consensus. If they split: low consensus with each perspective preserved.
4. Bias Transparency: Every source's training bias is labeled. An "alternative knowledge" model and a "mainstream institutional" model disagreeing tells you something about where the narrative fault line is.