In a topsy-turvy election cycle, people are concerned AI deepfakes will play a starring role

A robocall that cloned President Joe Biden’s voice in January is one of the latest examples establishing that artificial dupes of real people’s voices and images can—and will—be used to trick members of the public.

In a new survey, identity verification platform Jumio found that the phenomenon is never far from the public’s consciousness, especially when it comes to politics.

Of the more than 8,000 adults queried across the US, UK, Mexico, and Singapore, 72% reported worrying about deepfakes on a daily basis. Of the countries surveyed, Americans worried the least—just over half reported that deepfakes were a constant concern. Mexicans reported worrying the most; 89% of respondents said it’s always on their mind.

More than half of the respondents also agreed that their governments don’t do enough to regulate artificial intelligence, while 43% expressed “faith” in their government’s ability to do so.

The pervasiveness of AI also extends to perceived election integrity. Three-fourths of respondents across the world said they “worry about the potential for AI and deepfakes to influence upcoming elections,” including 72% of American participants.

While 60% of respondents expressed bravado about their ability to spot deepfakes—with young men communicating the highest confidence that they could detect AI manipulation—that assurance declined in the context of elections.

Only one in three Brits “think they could easily spot a deepfake of a politician,” and Americans expressed similarly low confidence. Jumio also found that the older a participant was, the less confident they were in their ability to detect a deepfake.