EP 548: Disney vs. Midjourney - Will Hollywood kill AI?
🎯 Summary
Podcast Summary: EP 548: Disney vs. Midjourney - Will Hollywood kill AI?
This episode of the Everyday AI Show focuses on the recent copyright infringement lawsuit filed by Disney and Universal against the AI image generator Midjourney. The host, Jordan Wilson, analyzes the implications of this high-profile case, particularly concerning the training data used by generative AI models and the liability shift towards end-users.
1. Focus Area
The primary focus is Generative AI and Copyright Law, specifically examining the legal battle between major Hollywood IP holders (Disney/Universal) and a leading independent AI image generator (Midjourney). Key themes include data scraping legality, visual copyright infringement, and the evolving Terms of Service (ToS) within the AI industry.
2. Key Technical Insights
- Visual Plagiarism Evidence: The lawsuit hinges on visually compelling evidence where Midjourney outputs (e.g., Stormtroopers, Minions, Shrek) are nearly identical to copyrighted material, even from generic prompts, suggesting direct reproduction rather than abstract inspiration.
- Training Data Sourcing: The core technical dispute revolves around Midjourney’s alleged practice of scraping vast amounts of internet data, including copyrighted works, without consent to build its model, which the studios label as a “bottomless pit of plagiarism.”
- Filtering Parity: The studios argue Midjourney possesses the technical capability to filter infringing content (as it does for nudity or violence) but allegedly refused to implement copyright blocks following initial warnings.
3. Business/Investment Angle
- Midjourney as the “Sacrificial Lamb”: The host posits that Disney and Universal are targeting Midjourney—a highly successful but smaller entity (projected $500M revenue in 2025)—instead of trillion-dollar giants like OpenAI or Google, viewing Midjourney as the most manageable “domino” to fall first in the creative AI space.
- Industry Precedent: A win for Disney could set a significant legal precedent, potentially forcing other AI creative startups to drastically alter their training methodologies or face similar legal action, signaling a unified Hollywood front.
- User Liability Shift: Midjourney’s updated Terms of Service (early 2024) explicitly shift 100% of the legal liability, including indemnification for legal fees, onto the user if copyright claims arise—a practice the host notes is unusual for major AI platforms.
4. Notable Companies/People
- Midjourney: The defendant, one of the earliest and most popular AI image generators, now facing a direct challenge to its training data practices.
- Disney and Universal (NBCUniversal): The plaintiffs, representing the consolidated power of Hollywood IP ownership, seeking an injunction against infringing outputs.
- OpenAI/Microsoft & Google: Mentioned as larger entities that studios are currently avoiding due to their deeper financial and legal resources, contrasting with the strategy against Midjourney.
- Jordan Wilson (Host): The host, drawing on his background as an investigative reporter, analyzed the 130-page complaint and tracked Midjourney’s ToS changes over time.
5. Future Implications
The outcome of this lawsuit is seen as a potential “first big domino” that could reshape how image generation models are trained and deployed. If Midjourney loses, it could force a massive overhaul in data acquisition practices across the industry, potentially slowing down the rapid pace of visual AI development unless models can prove their outputs are transformative enough to qualify as fair use. The episode emphasizes that users must be acutely aware of the ToS of the tools they use, as liability is increasingly being pushed onto them.
6. Target Audience
This episode is most valuable for AI Professionals, Legal/IP Counsel, Tech Investors, and Creative Industry Leaders who need to understand the immediate legal risks associated with using or developing generative AI tools based on scraped internet data.
🏢 Companies Mentioned
💬 Key Insights
"The AI generates copyrighted characters, even when users give generic prompts that don't specifically name the character."
"The third one that I saw was most important: infringement without specific prompts."
"essentially, in early 2024, Midjourney has said, "No, if there's ever a lawsuit, if there's ever any IP, anything, it's on you.""
"The studios are alleging that Midjourney built their AI by illegally copying vast amounts of copyrighted material."
"Midjourney CEO admitted that they pull off all the data it can and they never sought any copyright content holders' consent to copy and exploit their works."
"Number four... there was a willful, this record, and continued infringement because the studios alleged that Midjourney has the technical ability to block infringing content, which they use for other filters such as nudity or violence, and that the studios did send Midjourney a cease and desist letter, but that they reportedly refused to implement any copyright protections..."