On an initial glance, a video that’s going viral on social media looks like a typical 1990s TV commercial for a children’s toy set of a tropical island: there are palm trees, a waterfall, and — wait a minute — there’s also a secret massage room, and even an “Orange Man” action figure who looks exactly like Donald Trump and says “don’t release the files” in an ominous voice.
“Kick back with two chill old dudes!” enthuses a manic-sounding narrator. “The only island with real working spy cams hidden in secret rooms!”
It’s a dark parody poking fun at deceased financier Jeffrey Epstein, in other words, who’d been accused of sex trafficking underage girls to his Caribbean island before his death in 2019. And it was all cooked up in OpenAI’s AI video generator Sora 2, which was released this week with a TikTok-style app for sharing AI-generated clips.
The satire is part of an avalanche of surreal videos churned out by users of Sora 2 since its release, including a now-viral clip of OpenAI CEO Sam Altman stealing GPUs from Target shelves.
The Epstein video presages tough decisions facing OpenAI about how it’ll moderate use of the brand new platform. On a certain level, the clip has elements of the sort of cutting satire you might see on an edgy sketch comedy show; on the other hand, it’s unquestionably making fun of real-world child abuse. And as a practical matter, it shows that the platform has limited guardrails against disparaging living public figures — including the infamously thin-skinned Trump, who currently holds immense regulatory power over the AI industry.
Unsurprisingly, the fake ad immediately ginned up controversy.
“This is not the AGI,” one person wrote on X, tagging Altman. “As a parent this leaves such a bad taste. Terrible.”
“Orange man bad?” another X user wrote. “OpenAI will have a lawsuit on their hands pretty soon, or at least get stopped payments from any government work.”
The issue of moderating sensitive topics is a major one for OpenAI. It was already facing explosive controversy over a series of deaths linked to the use of its platform, including the suicide of a teen boy. Now that it’s expanding operations into social media, it’s inviting the type of nonstop scrutiny over what type of content it’ll allow that have haunted platforms ranging from Twitter to Facebook.
OpenAI’s Sora 2 safety document obliquely addresses those concerns, saying it has filters that “block unsafe content before it’s made — including sexual material, terrorist propaganda, and self-harm promotion.”
But as the Epstein video shows, it’s easy for content to fall into a deeply gray area.
“Epstein’s Island is associated with highly sensitive, illegal, and exploitative activities, including human trafficking and abuse,” one X user wrote in a critique below the video, spelling out the stakes. “Content related to such a topic is deeply inappropriate and raises serious ethical, legal, and social issues.”
As for the appearance of Trump’s plastic avatar in the video, that appears to violate Sora 2’s protocols; the app is supposed to “block depictions of public figures,” as spelled out in the safety document.
One thing’s for sure: there’s going to be a nonstop torrent of objectionable material like this for OpenAI to contend with — as soon as any company steps into the world of social media, that’s just the name of the game.
More on OpenAI’s Sora: OpenAI’s Sora 2 Generates Realistic Videos of People Shoplifting