Don’t come after the king: OpenAI restricts Sora after MLK deepfake videos spark backlash

OpenAI has temporarily suspended its AI video generator Sora from creating deepfake videos of Martin Luther King Jr., following a request from the civil rights leader’s estate. The move comes after “disrespectful” content depicting Dr. King was generated and shared widely online, including altered versions of his famous “I Have a Dream” speech and fabricated scenarios showing him in offensive situations.

What happened: OpenAI paused the ability to create AI videos of Dr. King while it works to strengthen guardrails for historical figures, though users can still generate content featuring other deceased celebrities and public figures.
• The decision followed complaints from Bernice A. King, Dr. King’s daughter, who publicly asked people to stop creating AI videos of her father.
• Similar requests came from Zelda Williams, daughter of Robin Williams, who asked people to stop sending her AI-generated videos of her late father.
• Videos featuring other historical figures like President John F. Kennedy, Queen Elizabeth II, and Professor Stephen Hawking continue to circulate on the platform.

The problematic content: AI-generated videos of Dr. King included deeply offensive material that distorted his legacy and message.
• Some clips showed altered versions of his “I Have a Dream” speech, with one reportedly featuring racist sounds, according to the Washington Post.
• Other videos depicted fabricated scenarios showing Dr. King and Malcolm X fighting each other.
• The content was shared both on Sora’s platform and across various social media channels.

What experts are saying: AI ethicists view the temporary pause as progress but argue OpenAI should have implemented protections from the start.
• “It plays too closely with trying to rewrite aspects of history,” AI ethicist Olivia Gambelin told the BBC, warning that such deepfakes pose dangers beyond disrespect.
• Gambelin criticized OpenAI’s “trial and error by firehose” approach to rolling out the technology without proper safeguards.
• She emphasized that creating deepfakes of deceased historical figures threatens people’s ability to distinguish between real and fake content.

OpenAI’s response: The company acknowledged the tension between free speech and protecting the likenesses of historical figures and their families.
• “While we believe there are strong free speech interests in depicting historical figures,” OpenAI said, families and estates “should have control over their likenesses.”
• The company allows “authorized representatives or estate owners” to request that their likeness not be used in Sora’s “cameos” feature.
• OpenAI stated it has built “multiple layers of protection to prevent misuse” and is in “direct dialogue with public figures and content owners” to gather feedback on desired controls.

Why this matters: The controversy highlights broader concerns about deepfake technology’s potential to spread disinformation, discrimination, and abuse, particularly when applied to revered historical figures whose legacies remain central to ongoing social and political discourse.