OpenAI blocks MLK Jr. videos on Sora after ‘disrespectful depictions’

OpenAI has blocked users from making videos of Martin Luther King Jr. on its Sora app after the estate of the civil rights leader complained about the spread of “disrespectful depictions.”

Since the company launched Sora three weeks ago, hyper-realistic deepfake videos of King saying crude, offensive or racist things have rocketed across social media, including fake videos of King stealing from a grocery store, speeding away from police and perpetuating racial stereotypes.

Late on Thursday, OpenAI and King’s estate released a joint statement saying AI videos portraying King are being blocked as the company “strengthens guardrails for historical figures.”

OpenAI said it believes there are “strong free speech interests” in allowing users to make AI deepfakes of historical figures, but that estates should have ultimate control over how those likenesses are used.

The Sora app, which remains invite-only, has taken a shoot-first, aim-later approach to safety guardrails, which has raised alarms with intellectual property lawyers, public figures and disinformation researchers.

When someone joins the app, they are instructed to record a video of themselves from multiple angles and record themselves speaking. Users can control whether others can make deepfake videos of them, which Sora calls a “cameo.”

But the app allowed people to make videos of many celebrities and historical figures without explicit consent, enabling users to create fake footage of Princess Diana, John F. Kennedy, Kurt Cobain, Malcolm X and many others.

Kristelia García, an intellectual property law professor at Georgetown Law, said for OpenAI to only act after King’s estate complained is consistent with the company’s “asking forgiveness, not permission” approach.

“The AI industry seems to move really quickly, and first-to-market appears to be the currency of the day (certainly over a contemplative, ethics-minded approach),” García told NPR in an email.

She noted how right-to-publicity and defamation laws vary by state and may not always apply to deepfakes, meaning there could be “little legal downside to just letting things ride unless and until someone complains.”

While the ability to control how one’s likeness depends on where someone’s estate is based, some states have strong protections, like California, where heirs to a public figure, or their estate, own the rights to likeness for 70 years after a celebrity’s death.

In the days after the Sora app was released, OpenAI CEO Sam Altman announced changes to the app providing rights holders the ability to opt into their likenesses being depicted by AI, rather than such portrayals being allowed by default.

Still, the families of some deceased celebrities and public figures have criticized OpenAI for allowing depictions of vulgar, unflattering or incriminating behavior.

After videos of Robin Williams flooded social media feeds, Zelda Williams, the late actor’s daughter, asked the public to stop making videos of her father. “Please, just stop sending me AI videos of my dad,” she wrote in an Instagram post, adding that “it’s NOT what he’d want.”

Bernice King, the civil rights leader’s daughter, agreed, writing on X: “Please stop.”

Hollywood studios and talent agencies have also expressed concern that OpenAI unveiled the Sora app without receiving consent from copyright holders.

It’s an approach similar to how the company has developed ChatGPT, which sucked up droves of copyrighted content without approval or payment before eventually striking licensing deals with some publishers. The approach has sparked a wave of copyright lawsuits.