#143. 🤯 Disney attacks Midjourney; Getty retreats from Stability

This is educational material and does not constitute legal advice nor is any attorney/client relationship created with this article, hence you should contact and engage an attorney if you have any legal questions. No warranties, express or implied, are made with respect to its accuracy. Information contained herein, or information relied upon, is subject to change without notice.

A seismic shift in the AI-copyright saga dropped on June 25, 2025, when Getty Images abruptly scaled back its high-stakes lawsuit against Stability AI in the UK High Court, mere hours after US federal judges ruled in favor of Anthropic and Meta in separate author-led cases.

Those rulings declared AI training on copyrighted material “spectacularly transformative,” cementing fair use protections under Section 107 of the Copyright Act. The decisions dissected the four-factor test — purpose, nature, amount, and market impact — and found that AI outputs, when sufficiently novel, don’t infringe.

Getty’s retreat from its core claims, which accused Stability AI of scraping 12 million images to train Stable Diffusion, likely stems from these US precedents signaling dim prospects for training-based infringement suits, compounded by UK jurisdictional snags over proving where Stability’s training occurred.

Meanwhile, the Disney v. Midjourney case, filed June 11, 2025, looms large. Unlike Getty’s broad image-scraping claims, Disney and Universal zero in on Midjourney’s outputs — near-identical renditions of Darth Vader, Elsa, and the Minions — arguing they’re not transformative enough for fair use.

Precisely on point, the Marin County Bar Association just published my article Why AI Training Should Fall Outside Copyright’s Domain: A Legal Analysis. Available both as an interactive Issu magazine print and plain HTML versions, with validation and guidance from Stanford’s Mark Lemley, it directly anticipates this tide change in courts’ perceptions on training AI on scraped copyrighted works. I argue that AI training in general is transformative and thus fair use, and that the burden of infringement should lie on volitional act of the end user.

These developments are a triumph for AI innovation, affirming that training on copyrighted works can be fair use, while signaling that blatant output mimicry or data piracy won’t slide. As the Disney v. Midjourney and NYT v. OpenAI cases unfold, expect more clarity on balancing creators’ rights with AI’s transformative potential.

Looking for past newsletters? You can find them all here.