# Interactive IP, AI & Licensing Reference Cases **Purpose:** a research reference for the Interactive & Narrative Writing path in *50 Ways to Leave Another*. These cases illuminate recurring questions when narrative work moves into games, interactive media, AI-assisted production, music generation, or licensed audiovisual experiences. This is an educational research index, not legal advice. Court rulings are fact-specific; some matters remain on appeal or in active litigation. Status is presented as of **September 25, 2026**. ## Why this belongs in the interactive-writing curriculum The REN's Rhapsody work extends the project's existing rights question: > **Create from the rights reality, not the rights assumption.** For interactive narrative, the useful question is rarely only “Is this copyrighted?” It is usually: - What exactly is being used? - Who owns or controls each layer? - How was the material obtained? - What permission exists for this particular use? - Does it cover interactive, promotional, commercial, derivative, or AI-assisted use? - What happens if the contemplated use changes after the prototype works? REN's model is especially useful because it separates **concept → reference → rights-dependent → cleared** rather than treating inspiration as permission. --- ## Reference cases and precedents ### 1. Thomson Reuters v. Ross Intelligence — training data, copying, and fair use **What happened:** Ross wanted to build a competing AI legal-research product. Thomson Reuters declined to license Westlaw material. Ross then obtained “Bulk Memos” built from Westlaw headnotes and used them in training. In February 2025, the Delaware federal court granted Thomson Reuters summary judgment on substantial portions of the copyright claim and rejected Ross's fair-use defense. **Why it matters here:** This is a clean example of the difference between **wanting access to a work, being refused a license, finding another route to the material, and then relying on fair use**. It is useful for teaching provenance and permission chains in AI-assisted narrative systems. **Primary source:** [U.S. District Court for the District of Delaware decision](https://law.justia.com/cases/federal/district-courts/delaware/dedce/1%3A2020cv00613/72109/770/). --- ### 2. Bartz v. Anthropic — training use versus source acquisition **What happened:** Anthropic faced claims that copyrighted books were used to train Claude and that pirated copies were stored in a central library. A 2025 ruling treated the training use as fair use on the record before the court while finding separate problems with the acquisition/storage of pirated books. In July 2026, a federal judge approved a **$1.5 billion settlement** covering the class claims. **Why it matters here:** It demonstrates a critical distinction: **the legal treatment of a model's training use and the legal treatment of how source material was acquired or stored are not necessarily the same question**. A project should document both provenance and contemplated use. **Source:** [Reuters, July 20, 2026](https://www.reuters.com/world/us/us-judge-approves-anthropics-15-billion-settlement-copyright-lawsuit-2026-07-20/). --- ### 3. Getty Images v. Stability AI — training, outputs, trademarks, and jurisdiction **What happened:** Getty Images sued Stability AI in the UK over Stable Diffusion. In November 2025, the High Court found trademark infringement in specified circumstances involving Getty/iStock watermarks but rejected Getty's secondary copyright-infringement claim in that proceeding. The training-related copyright questions were constrained by jurisdictional and procedural issues, and appeal activity followed. **Why it matters here:** An interactive project can cross several rights categories at once: **copyright, trademark, database rights, output behavior, and jurisdiction**. “The asset is online” is not a rights analysis. **Primary source:** [Courts and Tribunals Judiciary — November 4, 2025 judgment](https://www.judiciary.uk/judgments/getty-images-v-stability-ai/). --- ### 4. Andersen et al. v. Stability AI et al. — artists, training images, style, and attribution **What happened:** Visual artists brought claims concerning the use of their works in AI-image training and the ability of models to produce images associated with particular artists. The litigation has proceeded through multiple rulings and discovery stages rather than producing one simple rule for “style.” **Why it matters here:** This separates **copyright in a particular work** from broader questions about style, artist identity, attribution, and model behavior. It also illustrates why a provenance record should identify actual source works instead of relying on vague claims about “inspiration.” **Primary source:** [N.D. Cal. proceedings](https://law.justia.com/cases/federal/district-courts/california/candce/3%3A2023cv00201/407208/316/). --- ### 5. Disney + Universal v. Midjourney — recognizable entertainment IP in generative outputs **What happened:** Disney and Universal sued Midjourney in 2025, alleging unauthorized use of their copyrighted entertainment properties in the company's AI image-generation system and outputs. **Why it matters here:** This is the entertainment/IP analogue to the artist cases: a narrative project can contain **characters, worlds, logos, costumes, creatures, dialogue, and visual signatures** that create separate rights questions. An AI tool does not make those underlying IP questions disappear. **Source:** [Associated Press, June 11, 2025](https://apnews.com/article/722b1b892192e7e1628f7ae5da8cc427). --- ### 6. AFM v. Universal Music Group / Warner Records / Atlantic — licensing can create a second rights layer **What happened:** In June 2026, the American Federation of Musicians sued major record labels, alleging that licensing arrangements involving AI companies Suno and Udio failed to comply with contractual obligations owed to musicians. The labels' licensing relationships with AI companies therefore became a separate question from the underlying AI copyright litigation. **Why it matters here:** This is highly relevant to REN's Rhapsody. **A label can have authority to license catalog material while performers, unions, publishers, composers, or other participants may have additional contractual or statutory interests.** A “license obtained” checkpoint should identify what layer was actually cleared. **Sources:** [Reuters, June 5, 2026](https://www.reuters.com/legal/litigation/musicians-union-sues-record-labels-over-ai-licensing-2026-06-05/); [AFM litigation summary, August 21, 2026](https://www.afm.org/2026/08/summary-of-the-afm-v-universal-warner-and-atlantic-new-use-litigation/). --- ### 7. Suno + Warner Music Group + BMG — a positive licensing counterexample **What happened:** In September 2026, Suno announced new AI music models through collaborations with Warner Music Group and BMG, with participating artists' licensed works used as part of the model/product relationship. **Why it matters here:** Not every AI/IP story is litigation. This is the other branch of the curriculum: **rights holders and AI companies can structure permission, participation, and commercial use explicitly**. For REN's Rhapsody, this supports designing the creative architecture now while artist-specific implementation waits for a documented rights path. **Source:** [Reuters, September 9, 2026](https://www.reuters.com/legal/litigation/suno-releases-new-ai-music-models-partnership-with-warner-music-bmg-2026-09-09/). --- ### 8. Lionsgate + Runway — licensed AI training as a studio collaboration **What happened:** In 2024, Lionsgate and Runway announced a partnership centered on a customized AI model trained on Lionsgate's proprietary film and television portfolio for use by Lionsgate and its creative talent. **Why it matters here:** This is a practical precedent for the question **“What would a rights-aware AI workflow look like if the rights holder participates?”** It contrasts with cases where training data was disputed. **Source:** [Lionsgate announcement, September 18, 2024](https://investors.lionsgate.com/news-events/news/news-details/2024/Runway-Partners-with-Lionsgate-in-First-of-its-Kind-AI-Collaboration/default.aspx). --- ## How these map to REN's Rhapsody | REN question | Reference lesson | |---|---| | Can we prototype before music rights are resolved? | Build mechanics and creative language from original, cleared, or placeholder material; don't convert a reference into an assumed license. | | Does permission for a recording cover the composition? | Treat rights as layers; identify the actual grant rather than “music rights” as one checkbox. | | Does permission for music cover interactive/game use? | Match permission to the actual medium, territory, term, promotion, monetization, and derivative/interactive uses. | | Can AI be used because the output is new? | Provenance, training/source use, output similarity, and commercial use can present different legal questions. | | Can a rights holder license AI use? | Yes, structured licensing and participation models exist; the permission needs to be documented. | | Can a label's permission settle every participant's issue? | Not necessarily; contracts, union agreements, publishing, performer interests, and other layers may remain. | | What should the website show while clearance is unresolved? | Separate **reference/inspiration** from **rights-dependent public use** and label the boundary. | ## The production rule **Verifiable relationship → rights matrix → creative mapping → prototype → clearance → final integration.** That is the reusable lesson. The case studies are not a list of “AI is legal” or “AI is illegal” precedents; they are examples of different facts producing different rights questions. ## Research boundary This page is a dated research index, not a substitute for legal review. Case status can change through appeals, settlements, later orders, contractual developments, or legislation. Before relying on a case for a transaction or production decision, read the primary decision and have project-specific counsel review the actual facts and proposed use.