# AI Training Data and Copyright Lawsuits and claims about what AI models are trained on: pirated "shadow libraries", scanning of purchased books, fair use, statutory damages, and AI companies' own statements about training data. It also covers the policy debate over copyright terms. Related topics: [[Topics/LLMs/LLMs|LLMs]] · [[Topics/Grokbot/Grokbot|Grokbot]] · Account: [[Users/@elonmusk/@elonmusk|@elonmusk]] · Index: [[X Research]] ## Post analyses filed | Post | Analysis | Verdict | |---|---|---| | [@GivnerAriel, Fri Nov 6, 2026, 1:22 PM CST](https://x.com/GivnerAriel/status/2118786000625488272) | [[Users/@GivnerAriel/Post Analyses/Post Analysis - GivnerAriel on Chicken Soup for the Soul v xAI (2026-11-06)\|Chicken Soup for the Soul v. xAI]] | ✅/🟡 Accurate recap of the real N.D. Cal. complaint (3:26-cv-13188, Nov 4, 2026). Musk quotes, $20B raise and SpaceX deal check out; 600M users is X+Grok combined, the 'two other publishers' are one publisher plus a rights-holder with authors, and the Grok 'vacuumed up' quote is unverified | ## Case tracker | Case | Court / No. | Filed | Status (Nov 8, 2026) | |---|---|---|---| | Chicken Soup for the Soul, LLC v. SpaceX Super Intelligence Corp. et al. (f/k/a X.AI) | N.D. Cal. 3:26-cv-13188 | Nov 4, 2026 | Complaint only. 4 counts (shadow-library copying, scanning purchased books, contributory, §1202(b) CMI). Individual, not a class. Jury demanded. | | Cognella, Inc. v. SpaceX Super Intelligence Corp. | N.D. Cal. 3:26-cv-13189 | Nov 4, 2026 | Same counsel (Stris & Maher; Freedman Normand Friedland) and theory | | Cambronne, Inc. et al. (Barretta, Adams, Sacks, Kochin) v. SpaceX Super Intelligence | N.D. Cal. 3:26-cv-13187 | Nov 4, 2026 | Cambronne claims *Bad Blood* copyright. Initial CMC Feb 9, 2027 | | Bartz v. Anthropic (background) | N.D. Cal. (Alsup, J.) | 2024 | 2025 ruling: scanning purchased books for training was fair use; pirated library copies were not. Class settlement ≈$1.5B ◻ | ## Musk statements cited in the complaints | When (CT) | Statement | Source | |---|---|---| | Wed Nov 29, 2023 (DealBook Summit) | "…a huge lie… These AIs are all trained on copyrighted data. Obviously." · "By the time these lawsuits are decided we'll have Digital God." | Techdirt; Wccftech | | Fri Apr 11, 2025, 6:40 PM CDT | "@jack I agree" (to Dorsey's "delete all IP law") | [X](https://x.com/elonmusk/status/1910840422789763511) | | Mon Jul 27, 2026, 3:48 PM CDT | "I've asked the SpaceXAI team to preserve any rare books in a library and scan them the hard way vs just cutting off the spine and scanning" | [X](https://x.com/elonmusk/status/2081844165881594362) | ## Notes - **Marshall's position** (reply, [Nov 7, 2026](https://x.com/marshallpittman/status/2119108203648602202)): copyright should be short, enough for living authors to earn, then pass to the commons. For comparison, the current US term is life + 70 years, while the 1790 Act gave 14 years plus a 14-year renewal ◻. - A chatbot's statements about its own training data ("vacuumed up whatever was out there") are weak evidence. Models don't have reliable access to their training records.