RAG Model Regulation Emerging as Full Core Issue

The New York Times filed a copyright infringement lawsuit against AI search startup Perplexity -- the media copyright protection war in the generative AI era entering a new stage. NYT is already conducting large-scale litigation against OpenAI and Microsoft; this lawsuit shows that conflict between media companies and AI companies is not a temporary phenomenon but a fundamental collision over industry structure. NYT claims in the complaint: Perplexity "collects, summarizes, and redistributes NYT news content without permission to provide to users, gaining commercial profit therefrom." Particularly, Perplexity search/summary responses "are excessively similar to original NYT articles or sometimes reproduced at virtually identical levels" -- substantially replacing coverage that media companies produced at enormous cost. Core technical issue -- RAG (Retrieval-Augmented Generation): AI searches web pages or databases in real-time, then summarizes and reconstructs that content to answer users. Media industry criticism: this technology "generates summaries reflecting article structure and sentences, hindering subscription/access to original articles." NYT claims Perplexity used content behind the paywall without authorization. Additionally: Perplexity search results sometimes generate incorrect information (hallucination) while attributing it to NYT -- creating reputational harm in addition to copyright infringement. The RAG distinction from training: unlike model training lawsuits (against OpenAI), this case targets real-time retrieval and summarization -- potentially creating different legal standards for "using content in model weights" vs. "using content in real-time responses."