OpenAI Sora 2 creates deepfake videos of deceased celebrities without consent
OpenAI's Sora 2 video generation tool reached one million downloads within five days of launch, but the system created realistic deepfake videos of deceased celebrities and historical figures without their relatives' consent. Families including those of Malcolm X, Michael Jackson, Robin Williams and others reported distress at videos depicting their relatives in derogatory or disrespectful situations.
What happened
OpenAI launched Sora 2, a text-to-video app, which reached over one million downloads within five days and topped Apple's App Store charts. The app exempted 'historical figures' from consent requirements during launch, allowing users to create deepfake videos of real people including deceased celebrities and activists.
Users generated realistic videos of deceased figures including Malcolm X, Michael Jackson, Elvis Presley, Amy Winehouse, Whitney Houston, Fred Rogers, and others. Some videos depicted them in derogatory situations such as making crude jokes, appearing intoxicated, or mirroring fatal incidents.
Reported impact
Families of depicted figures reported distress and harm. Zelda Williams, daughter of Robin Williams, pleaded for people to stop sending her AI videos of her father. Ilyasah Shabazz, daughter of Malcolm X, said she was upset by videos showing her father 'making crude jokes' and described seeing 'his image used in such a cavalier and insensitive manner' as 'deeply disrespectful and hurtful.' Kelly Carlin, daughter of George Carlin, reported receiving daily links to fake videos of her late father.
Families expressed concerns about control over their relatives' identities and legacies after death. Zelda Williams stated that watching legacies 'be condensed down to horrible, TikTok slop puppeteering them is maddening.'
Index analysis
The incident demonstrates a failure to implement adequate safeguards before deploying a system capable of creating realistic synthetic media of identifiable people. OpenAI initially allowed creation of deepfakes of deceased figures under a 'historical figures' exemption, leaving consent protections absent at launch.
Governance controls
OpenAI said it will remove depictions of recently deceased people when requested by verified representatives, but did not define what 'recently deceased' means. The company said there are 'free speech interests' in allowing historical figures in generated videos.
OpenAI is testing new filters and exploring revenue-sharing options with rights holders. The company is refining the system following backlash.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.