Anthropic didn’t just leak its code, it revealed a trap for its competitors. When a simple human error led to an accidental leak of Claude’s architecture, it exposed an unannounced feature with its anti-distillation. As SCSP AI Engineer Vesa Xërxa explains, competitor AI models have been creating thousands of fake accounts to scrape Claude’s answers and train their own systems (essentially using Claude as a free teacher). But Claude knows, and it fights back. If it detects a competitor scraping its data, it intentionally injects false information to poison their training model. #ArtificialIntelligence #TechNews #Anthropic #Claude #ainews