Seedance 2.0: What risks and what liability for content created using AI?

Since last February, a new AI tool has been making a big splash: Seedance 2.0, a video creation model developed by a subsidiary of ByteDance, TikTok's parent company. This Chinese tool sparked outrage in Hollywood after a flood of videos surfaced on social media showing alternative endings to cult movies and series such as Titanic, Game of Thrones, or Stranger Things, all having in common that they were generated with Seedance 2.0. The most viral video shows a showdown between Brad Pitt and Tom Cruise over the Jeffrey Epstein affair.

Considered the most advanced AI video generation model, Seedance 2.0 reportedly surpasses OpenAI’s Sora 2 and Google’s Veo 3.1. In recent weeks, Hollywood giants have united against its global launch, citing clear copyright infringement. As a result, ByteDance has temporarily suspended the scheduled global launch. While these videos are fascinating for their realism, they raise a major legal question: can a user be held liable when using an AI tool trained on protected content to generate, intentionally or not, content that reproduces a work or the likeness of a real person?

There are two levels of liability related to the use of this type of AI tool: liability associated with training AI tools on protected content, on the one hand, and liability associated with creating and distributing content generated by these same tools that unlawfully reproduces the likeness of a real person or a pre-existing work, on the other hand. CAN AN AI USER BE HELD LIABLE SOLELY FOR USING AN AI TOOL?

AI tools, such as Seedance 2.0, could not have emerged or developed so rapidly without massive data collection conducted without prior authorization or compensation for the individuals concerned. From the outset, the publishers of these tools built their models and systems in disregard of copyright and third-party personality rights.

In recent years, lawsuits opposing AI giants and rights holders—such as independent authors, record labels, or news publishers—have multiplied, particularly in the United States.

All of these disputes share a common goal: attempting to answer the following question: did the AI giants have the right to create their systems and models without asking for authorization or compensating authors/artists?

None of these lawsuits target a simple user of an AI tool solely for having used an AI tool designed in violation of third-party rights.

Consequently, there is virtually zero risk that you, as a user of an AI tool, will be held liable solely because you use the AI in question.

However, it is advisable to remain vigilant regarding the results generated with these AI tools before using them commercially. CAN A USER BE HELD LIABLE IN THE EVENT OF REPRODUCING PROTECTED CONTENT USING AI TOOLS?

A user who reproduces third-party works or trademarks using AI tools risks being subject to copyright infringement actions or right of publicity/image rights violation claims.

THIS IS TRUE:

■ even if your reproduction of movie clips or an artist’s image is only PARTIAL, or if it constitutes a DERIVATIVE adaptation of a tiny portion of your favorite shows. Thus, to create a derivative work, you must obtain prior authorization from the rights holder of the primary work.

■ even if you did not mean to obtain a result that represents a work or the likeness of a real person because you believe you prompted "ethically" (meaning without mentioning titles, works, or artists' names in order to draw inspiration from/plagiarize their works/images).

In matters of copyright infringement and violation of image rights, there is NO principle of "good faith." A user cannot be "excused" for unlawfully generating content that infringes upon the rights of third parties simply by proving that they were unaware of their existence.

Thus, for example, if the result you generated through an "ethical" creative process still gives the unsettling impression of being an adaptation of Matrix or depicting someone who looks like Brad Pitt, and you exploit this result, you will be held liable for a rights violation.

It is therefore your responsibility to ensure, prior to distribution, that the results generated with an AI do not infringe upon the rights of third parties.

There are several exceptions to this principle under French law, to be handled with caution:

■ when a person is deceased, it is possible to reproduce their voice or image without necessarily having to obtain the consent of their heirs/rightsholders. Note that this is not true everywhere in the world; it depends on local law.

■ it is possible to reproduce works that have "fallen into the public domain," that is, by artists who have been deceased for more than 70 years. Note that here too, the duration of copyright is not the same in every country in the world.

■ it is possible, under certain conditions, to create humorous, parodic, and satirical videos of fictional characters and famous personalities.

To further explore these issues, I refer you to the articles "AI & Copyright: what are the risks with parody?" (published in October 2025) and "Studio Ghibli-style" Images: can a style be legally protected? (published in May 2025). OUR LEGAL ADVICE

Before any exploitation of content generated using AI tools, it is essential to conduct an anti-plagiarism check, for example through a reverse image search, in order to identify potential pre-existing images or works that are similar or identical to the created content.