New AI Tool Makes Faking Reality Frighteningly Easy

They were all created with OpenAI’s Sora 2 video app, released Sept. 30. And they’re pretty darn good. And that could be pretty darn bad.

The software allows any user to create a shockingly convincing video with just text prompts – no programming or other tech skills needed. RIP “video or it didn’t happen.”

Yes, Sora typically embeds a Sora watermark that flashes at some point in these new artificial intelligence-generated videos. But that’s created a Sora watermark removal industry.

The company says its text-to-video app will only depict real people with their consent. But they exempted “historical figures.” Cue Sora videos of John F. Kennedy joking about Turning Point founder Charlie Kirk’s assassination and of late comic Robin Williams.

“To watch the legacies of real people be condensed down to ‘this vaguely looks and sounds like them so that’s enough,’ just so other people can churn out horrible, TikTok slop puppeteering them is maddening,” Williams’ daughter Zelda said in an Instagram post.

I observed in this column before that I’m an AI worrier. It’s in my Gen X DNA. But this is AI doomer rocket fuel and raises a lot of moral, legal, economic and, yes, political questions about when and how we should be using rocket fuel.

One thing Sora (and its less-convincing counterparts at Meta and Google) has done, right out of the gate, is make disinformation extremely easy. Welcome to the “lyin’ eyes” phase of the Internet.

Or, as the New York Times put it: “Increasingly realistic videos are more likely to lead to consequences in the real world by exacerbating conflicts, defrauding consumers, swinging elections or framing people for crimes they did not commit.”

No one (I hope) is going to buy a video of a vaping squirrel. And I honestly enjoy some of the videos of good samaritans “cleaning” whales of barnacles even though I know they’re fake. But grainy security-camera-style nighttime footage of someone sabotaging a ballot box? Ugh.

This White House has already embraced digitally altered videos. But while no one is going to think Democratic House Minority Leader Hakeem Jeffries, who is Black, wore a sombrero to the White House, ruthless political campaigns are sure to play with Sora.

The company isn’t blind to the potential problems. Every video has not just a visible watermark but “invisible provenance signals” to identify whence it came, OpenAI said at launch.

The company said it would enforce consent-based likeness for ordinary users and safeguards for teens, and would attempt to “block unsafe content before it’s made – including sexual material, terrorist propaganda and self-harm promotion.”

OpenAI also says it will work with copyright holders who have complained about infringement.

I have said on social media for years that people don’t just have to be skeptical of online content. They have to be especially skeptical of content that reinforces their beliefs.

They’re just getting a lot harder to find. Caveat surfator! (Surfer beware!)