Published: Jan 19, 2026, 23:36 IST | Updated: Jan 19, 2026, 23:36 IST
A groundbreaking study from Stanford and Yale researchers exposes how leading AI models like Claude, Gemini, GPT-4.1, and Grok can reproduce copyrighted books near-verbatim with up to 96% accuracy, shattering claims they merely "learn" from data. The findings trigger urgent copyright alarms for publishers facing potential "digital shoplifting" via cheap API extractions.