Now AI fakes are fooling news outlets — and maybe AI pros?

• “Fake News” used to be a real news story you didn’t like.
• Now it can mean “we got duped by a bogus AI video from Sora or somewhere else.”
• Fox News fell for one. Did one of the world’s most high-profile AI experts get tripped up, too?

For years, we’ve heard warnings that computers would soon be able to crank out realistic fake videos — the kind that could fool anyone. Even people who should know better.

Looks like that time has arrived.

On Friday, Fox News ran a story headlined “SNAP beneficiaries threaten to ransack stores over government shutdown.” The story explained that “SNAP beneficiaries have expressed outrage on social media over the government shutdown that could affect their grocery benefits starting next month — with some even threatening to ransack stores if food stamp payments don’t go through starting Nov. 1.”

Then, after online commentators pointed out that at least some of the clips cited in the Fox story were AI-generated, Fox rewrote the story with a new headline: “AI videos of SNAP beneficiaries complaining about cuts go viral.” An editor’s note at the bottom of the story says, “This article previously reported on some videos that appear to have been generated by AI without noting that. This has been corrected.” Fox didn’t return a request for comment.

Fox wasn’t alone. Newsmax aired a similar segment.

It’d be tempting to make a connection here between AI slop and a conservative news ecosystem that’s particularly susceptible to online clips that confirm existing biases. And maybe that’s true!

But it turns out that lots of people can (seemingly) be fooled by AI clips — especially ones that tell them something they’d like to believe is true.

On Sunday, for instance, Yann Lecun, a Meta AI researcher who until recently was its most prominent, posted a video that looks like a New York City cop telling ICE agents to “back off, now.”

I’ve seen other real videos of local cops in conflict with ICE agents recently. But this one certainly seems like something you can generate on Sora, OpenAI’s video-making app. And it has some of the tell-tale signs of a Sora-esque video — primarily its short duration and stilted speech patterns.

Did Lecun think this one was real? Or was it a wry commentary? I’ve reached out to him to ask, but haven’t heard back. (Meta reps declined to comment, and LeCun didn’t reply to an email.) But after Threads commenters suggested he had posted a bogus clip, Lecun posted this inscrutable rejoinder.

In the old days, discussions about the rise of deepfakes and other AI trickery would end with a cry for industry, or government, to step in. But I don’t see that happening anytime soon: None of the big AI outfits seems particularly interested in putting meaningful guardrails on the outputs of their software. The Trump administration doesn’t seem inclined to push for that, either.

Meanwhile, Donald Trump himself, who has already learned to call anything he doesn’t like “fake news,” now seems to think that AI fakes may be an opportunity, as well as a problem.

For starters, he can actively use them for his own ends, even if he may not be aware he’s doing that.

But at a White House press conference earlier this fall, Trump started workshopping a new idea — if confronted with footage he doesn’t like, he’ll simply say a computer made it.

“If something happens really bad, just blame AI,” he said. “But also they create things. It works both ways.”

Read the original article on Business Insider

The post Now AI fakes are fooling news outlets — and maybe AI pros? appeared first on Business Insider.