It's hard to know what's real, it turns out.

Sometimes things are obvious. When it’s raining, there’s no doubt about whether it’s raining or not. A flat tire, in all likelihood, really is a flat tire. But much of the world is not directly verifiable in this way. Are there really millions of dead people collecting Social Security? Does China secretly control the Panama Canal? Was that video on TikTok actually Tom Cruise? How do you know for sure?

The rapid ascent of generative AI is only making things worse. Within the past few years, it’s been used to create political deepfakes from Moldova to South Africa to the United States. An AI-generated Imran Khan, Pakistan’s ex-prime minister, gave a speech while the real Khan was in jail. AI-generated evidence is now being introduced in divorce court, and last year AI was used to impersonate a company’s CFO and steal $25 million. According to one survey, deepfake incidents rose in 2024 by 300% in the United States, 1,625% in South Korea, 2,800% in China, and 3,000% in Bulgaria. Even the suggestion of an AI deepfake is now enough to cast doubt on anything, from the size of political rallies to legal culpability in criminal court. 

Related

Last year, in an attempt to curtail the spread of misinformation, Meta launched a campaign called “Know What’s Real,” encouraging users to verify things before sharing them. A few weeks later The Atlantic launched a

© Salon