Amartya Kumar

AI now writes product descriptions, vendor briefs and market research for eCommerce teams — and it invents sources while doing it. In our published audit of 30 AI-generated briefs (194 citations), only 26% of citations verified and 49% were broken; in a smaller 66-citation pilot, just 5 citations both resolved and fully supported their claim. In this talk you will learn a four-verdict check — verified, broken, mismatched, unverifiable — that catches fabricated AI claims before they reach a customer or a contract, run on any AI-generated document in about ten minutes per batch using free tools. You will leave with the check itself, a one-page routine your team can run before anything AI-written goes live, and the two numbers to quote in your next vendor conversation: 49% broken, 26% verified.

Amartya Kumar is the Founder of the Oryx Centre for Strategic Futures, an independent research initiative in Noida, India (Udyam-registered micro enterprise, UDYAM-UP-56-0189301; listed in the UN DESA Civil Society database). His research focuses on verifying AI systems: auditing AI-generated citations, adversarial review of agentic systems, and human decision gates for commerce AI. His work on claim verification is published via Zenodo. This would be his first speaking engagement.