Disney accuses Google AI tools of using copyrighted characters

Disney has accused Google of large-scale copyright infringement, alleging that the tech giant’s AI models generate content resembling characters from Disney films and franchises.

The claim was made in a cease-and-desist letter sent by Disney on December 10, as reported by Variety and Deadline.

Disney named Google’s Gemini, Veo, Imagen, and Nano Banana models, asserting that these tools are producing unauthorized works based on Disney’s intellectual property.

The company also said it had tried to engage with Google for months about safeguards against copyright infringement, but did not reach a resolution.

This follows previous actions by Disney, including warnings to Character.AI and a lawsuit against Midjourney over similar concerns.

The cease-and-desist notice comes a day before Disney announced a deal with OpenAI to allow Sora AI users to create videos using over 200 Disney-owned characters.

🔗 Source: The Verge

🧠 Food for thought

Implications, context, and why it matters.

Google’s public copyright controls may not extend to AI training datasets

In replying to Disney, Google pointed to Google-extended (a setting that lets sites say they don’t want pages used to train Google’s AI) and YouTube Content ID (a system that identifies plus manages copyrighted audio or video uploads) 1. Both let owners or rights holders limit scraping or manage uploads, yet neither reveals what went into AI training sets.
Disney wants a list of copyrighted works used to train Google’s models 2. That request signals it sees a gap during the training stage.
Google Cloud says, “We do not use customer data to train our models” 3. It has not said whether public Disney works fed Veo or Imagen 3.
Disney says it warned Google for months while infringement allegedly grew 2. That timeline hints at missing guardrails that could stop character lookalikes if training included copyrighted works.

AI operators need copyright detection middleware before deployment

App teams using Gemini or Veo can add Azure AI Content Safety 4. Its Protected Material Detection APIs scan text or code for known copyrighted content. Middleware is a software layer between the model and the user app. It can backstop model controls.
One open-source detector in recent research cut compute use by 10-30% with efficient API calls and layered checks 5. That lift puts pre-release screening within reach for startups.
Investors have a signal from Disney’s deals with OpenAI for authorized character use 2. Licensing hubs that broker IP access between studios and AI firms could see demand beyond post-generation filters.
A University of Turin study for the European Union Intellectual Property Office (EUIPO) reviewed copyright in generative AI service development 6. Rising EU focus on training data and outputs rewards teams that can document compliance.