I tried to prove I’m not AI. My aunt wasn’t convinced

In Netanyahu’s case, it didn’t help that his team used a fancy camera and filmed with a narrow depth of field, meaning a nice sharp foreground and soft blurry background – which is exactly how AI videos tend to look, Carrasco says. But by the time Netanyahu posted his coffee shop clip, our world was already saturated with fake content, raising suspicions that are hard to overcome under any circumstances. Netanyahu’s fingers, hallucinated or otherwise, are just the latest wave in an AI flood that’s been rising for years.

This chaos has a name: researchers call it the “liar’s dividend”. It’s expensive to prove something is real, but casting doubts is free. “People, including people in positions of power, can argue that genuine content – genuine evidence of them doing something – is fake,” says Samuel Woolley, chair of disinformation studies at the University of Pittsburgh in the US.

Politicians and others can use the spectre of AI as a shield, crying deepfake to dismiss real evidence. But it’s a double-edged sword, and that same atmosphere of distrust is coming back to bite them. “The very politicians that pushed for this lack of moderation are now, in many ways, paying the consequences,” Woolley says.

More like this:

• The number one sign you’re watching an AI video

• I hacked ChatGPT and Google in 20 minutes

• People are selling your home address online. This privacy tool will help

Back on the phone with my aunt Eleanor, reality was bending. It turns out she’d heard the codeword advice and she actually had one for her kids and her husband, but I wasn’t in the loop. “I’ve read a lot of stories like that, where they talk about voices being cloned from YouTube videos,” she said. “That concerns me. It’s terrifying.”