Why journalism is our best defence against confidently wrong AI
The New York Times reports that Trump posted 19 fake images and videos during his 2024 presidential campaign, from the outright absurd to the more plausible—such as a woman wearing a “Swifties for Trump” T-shirt. Since his election he’s posted 28 AI images and videos, including self-portraits as a king and as the pope. For anyone worried that AI-generated imagery might become a routine part of political influence operations, that horse has left the stable.
A recent study from fact-checking site NewsGuard found chatbots responding to queries about the news with falsehoods 35 per cent of the time—roughly double the rate of a year prior.
Newsguard explains that the rise comes from AIs being programmed to answer questions even if they are uncertain if an answer is correct. A year ago, AI bots would refuse to answer questions about the news 31 per cent of the time, while during this more recent round of tests the bots never refused to answer.