'The Office' Clips Show Chinese AI Being Used To Plunder…

The rapid advancement of generative AI has sparked significant controversy regarding the unauthorized use of Hollywood intellectual property. Following the viral success of a deepfake video depicting a fight between Tom Cruise and Brad Pitt, major studios have begun aggressive legal efforts to hold AI developers accountable for copyright infringement.

What happened

The incident involving Tom Cruise and Brad Pitt was produced using ByteDance’s Seedance 2.0 model, which prompted swift legal responses from companies like Disney and Netflix. The resulting cease and desist letters pressured ByteDance to improve its safety guardrails. However, the issue remains a global challenge as new models emerge. Recently, the MiniMax H3 model, developed in China, gained viral attention for allowing users to generate clips of popular characters from The Office using simple text prompts. Although MiniMax is not officially available in the U.S. or the European Union, third-party access has enabled users to bypass restrictions.

Currently, MiniMax is defending a lawsuit brought by Disney, Universal, and Warner Bros., which alleges that the company's technology "pirates and plunders" copyrighted works on a massive scale. The studios cited instances where MiniMax used AI-generated videos of characters such as Deadpool, Wolverine, and The Joker to promote its model on YouTube. While MiniMax has since scrubbed these videos and restricted access to its full model in the U.S. due to ongoing legal proceedings, researchers continue to find ways to generate unauthorized content. Tests by the startup VN.ai have shown that even with updated tools like ByteDance’s Seedream 5.0, "red teaming" techniques—such as adjusting pixel size parameters—can successfully bypass safeguards to create recognizable renderings of protected characters like Elsa and Buzz Lightyear. This ongoing cycle highlights the difficulty of policing intellectual property in an era of rapidly evolving generative AI.