Sora creates deepfakes of dead celebrities like Michael Jackson despite OpenAI policy

OpenAI’s Sora video generation app allows users to create AI deepfakes of deceased celebrities like Michael Jackson, Tupac Shakur, and Malcolm X, despite the company’s stated policy of blocking depictions of public figures. The policy exemption for “historical figures” raises questions about consent, misinformation, and the potential misuse of AI-generated content featuring dead celebrities who cannot approve their digital resurrection.

What you should know: OpenAI explicitly permits AI-generated videos of deceased public figures while blocking living celebrities unless they consent through the Cameos feature.

• Users have created disturbingly realistic deepfakes of Michael Jackson, Bob Ross, Tupac Shakur, and Malcolm X that accurately mimic their voices and facial expressions.
• Some clips even incorporate licensed music, adding another layer of potential copyright complexity.
• The company told PCMag: “We don’t have a comment to add, but we do allow the generation of historical figures.”

The big picture: Sora’s launch as OpenAI’s first social media platform positions it to compete with TikTok while navigating complex ethical and legal boundaries around AI-generated content.

• The invite-only iOS app launched Tuesday with features that could boost its appeal but also expose OpenAI to intellectual property lawsuits.
• Users can also create deepfakes of fictional characters like Nintendo’s Mario and SpongeBob SquarePants, potentially triggering copyright disputes.

Why this matters: The policy creates a concerning precedent where deceased individuals have no protection against AI manipulation, potentially enabling misinformation or content that contradicts their values.

• AI-generated clips could be mistaken for historical fact or used to spread false information about historical figures.
• The technology raises ethical questions about depicting deceased celebrities in scenarios they might not have approved of while alive.

Current safeguards: OpenAI has implemented limited protections while the technology remains in early access.

• All Sora videos currently include watermarks to identify AI-generated content.
• Living public figures can only be depicted with explicit permission through the Cameos feature.

Industry context: OpenAI’s approach contrasts with Google’s previous struggles with AI-generated historical imagery.

• Google faced controversy last year for producing anachronistic historical images, including an African American Founding Father and Asian soldiers in the 1943 German military.
• Google paused people generation for several months to make “technical improvements” after the backlash.