AI Applied: Covering AI News, Interviews and Tools - ChatGPT, Midjourney, Gemini, OpenAI, Anthropic Navigating the AI Legal Maze: Perplexity's Predicament
6 snips
Dec 9, 2025 The podcast dives into the legal turmoil surrounding Perplexity and its mounting lawsuits. It discusses retrieval augmented generation (RAG) and potential legal liabilities from handling verbatim articles. The hosts debate the implications of paywalls and the responsibility of pirate sites in the AI landscape. They also explore the ways Perplexity might bypass paywalls, and contrast it with Meta's strategy of forming partnerships with news publishers to ensure accuracy and mitigate legal risks. Tune in for insights into the intricate relationship between AI and the law.
AI Snips
Chapters
Transcript
Episode notes
RAG Makes Perplexity's Case Unique
- Perplexity's lawsuits differ because they're accused of using retrieval augmented generation (RAG) to store and serve news verbatim.
- That behavior crosses from fair-use summarization into direct reproduction that publishers can legally challenge.
Storing Verbatim Content Is Risky
- If Perplexity stores paywalled articles verbatim in a DB and returns them, that likely violates publishers' rights.
- The legal risk grows when scraped third-party sites rehost paywalled content that then enters training or retrieval corpora.
Scraping Creates Attribution Ambiguity
- Training sets can unintentionally include copied paywalled text via third-party reposts, creating attribution ambiguity.
- Blacklisting everything verbatim is impractical given how models ingest diverse scraped sources.