The recent broader deployment of OpenAI's Sora 2 video model and its associated social application has prompted public discourse on the ethical implications of AI-generated content, particularly regarding the likenesses of deceased individuals. Zelda Williams, daughter of the late actor Robin Williams, publicly expressed distress over AI-generated videos of her father, stating on Instagram, "Please, just stop sending me AI videos of Dad. Stop believing I wanna see it or that I’ll understand. I don’t and I won’t."
OpenAI's Sora 2, capable of generating highly realistic deepfakes, reportedly allows users to create videos of deceased public figures with fewer restrictions than for living individuals. The platform's policies require consent for generating living persons' likenesses, but deceased individuals, including historical figures like Martin Luther King Jr., Franklin Delano Roosevelt, and deceased celebrities such as Robin Williams and John Lennon, have appeared in user-generated content, according to reports from TechCrunch.
The legal framework surrounding the use of deceased persons' likenesses in AI-generated content remains an area of ongoing discussion. The Student Press Law Center indicates that it is not illegal to libel the deceased. OpenAI has not provided clear guidelines on its internal criteria for generating specific deceased figures; TechCrunch's tests indicated the model would generate Robin Williams but not former President Jimmy Carter (who died in 2024) or Michael Jackson (who died in 2009). Unlike living individuals, deceased figures currently have no mechanism within Sora's "cameo" feature to establish preferences for their digital representation.
The controversy surrounding Sora 2 extends to broader intellectual property concerns. Critics have accused OpenAI of adopting a permissive approach to content generation, which initially included copyrighted characters. The Motion Picture Association (MPA) has called for action on this issue, asserting that "well-established copyright law safeguards the rights of creators and applies here." OpenAI CEO Sam Altman initially indicated an opt-out policy for intellectual property owners but has since reportedly stated the company will reverse this position to implement granular opt-in copyright controls.
The capabilities of Sora, described as one of the most realistic deepfake-capable AI models accessible, underscore the evolving challenges in AI governance. While platforms like xAI are noted to have fewer guardrails, allowing for the generation of more controversial content, the ethical and legal precedents established by the deployment of models like Sora are expected to influence the wider adoption and regulatory landscape for generative AI across various industrial applications, from synthetic data creation to virtual prototyping.
