ByteDance tightens safeguards for Seedance after IP backlash

ByteDance is working to improve safeguards on its Seedance 2.0 video-generation tool after facing criticism over alleged intellectual property violations, according to the company.

The controversy involves claims that Seedance used copyrighted content, including videos of Hollywood celebrities, Disney characters, and Japanese manga figures, to train its AI models.

Disney has sent a cease-and-desist letter to ByteDance over alleged infringement involving its Marvel and Star Wars characters, while industry groups such as SAG-AFTRA and the Motion Picture Association have also raised concerns about potential copyright breaches.

Japan’s government has launched an investigation into ByteDance over possible copyright violations involving Japanese manga and anime characters.

ByteDance has not disclosed the data sources used for training Seedance 2.0 but previously stated it sources content ethically and legally and paused some personalization features following privacy concerns from users.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

This fight focuses on control and copyright

Disney’s cease-and-desist targets alleged infringement by Seedance 2.0. It also fits Disney’s push to control and license its intellectual property for AI, rather than block AI outright 1.
Disney has a multi-year, billion-dollar deal with OpenAI, the company behind ChatGPT. The agreement lets users generate images and videos using Disney IP under a licensing deal 1.
Disney has sent similar cease-and-desist letters to Google and Character.AI (a startup that makes chatbots based on fictional and real-world characters) over unlicensed use of Disney IP 1.

This points to higher costs for AI development

Paramount also sent ByteDance a cease-and-desist letter. It alleges infringement tied to franchises including “Star Trek” and “The Godfather” 2.
Major studios and Hollywood trade groups such as the Motion Picture Association are aligning. Their actions raise the pressure on AI companies that train models on copyrighted content without permission 2.
The market may split between models trained on licensed data and models that face lawsuits or missing content.
Well-organized, famous IP catalogs could become more valuable. They may serve as data sets for the next wave of generative AI tools.