Journalists used artificial intelligence to check the authenticity of images from the attempted assassination of former President Donald Trump in Pennsylvania on Saturday night — demonstrating both the potential and the limitations of AI technology in minimizing the spread of misinformation online.
AI tools offered by Seattle-based non-partisan nonprofit organization TrueMedia.org were used by reporters to assess images that spread on social media after the shooting, including a photo that appeared to capture a bullet in mid-air behind Trump’s head.
That picture was taken by New York Times photographer Doug Mills. But the tendency of social media users to post images without attribution can make it difficult to determine the origins of content, especially when news is breaking.
While cautioning that its results were experimental, TrueMedia.org correctly indicated that photo was likely not manipulated, according to Oren Etzioni, the organization’s founder.
In a similar way, responding to a query from another journalist, the TrueMedia AI tools also found little evidence of manipulation in what turned out to be an Associated Press photograph of Trump raising his fist after the incident. That picture also circulated widely on social media.
However, much of the unverified information that spread on social media platforms after the shooting was the digital equivalent of word-of-mouth rumors, messages that stoked conspiracy theories on all sides of the political spectrum.
AI tools assembled and developed by TrueMedia.org focus on images, video, and audio, commonly known as deepfakes.
“TrueMedia, as much as I would like it to be a silver bullet, is only a partial solution,” said Etzioni, a longtime computer scientist and AI specialist, and professor emeritus at the University of Washington Allen School of Computer Science.
Etzioni emphasized the important role of bona fide media organizations and fact-checkers to verify information, while also echoing the national leaders who have been calling on the country to wait for the results of investigations into the shooting before jumping to conclusions.
TrueMedia.org is “still a small part of a big issue,” said Etzioni, who is also the former CEO of the Allen Institute for AI in Seattle. “We have to, all of us, protect democracy, each in the way that we can.”
Separately, media verification service NewsGuard recently credited TrueMedia’s tools for helping it to identify 41 TikTok accounts that used “AI-generated narration to spread political misinformation at scale” — a growing phenomenon that NewsGuard described as “the rise of the TikTok AI content farm.”
TrueMedia.org, which was announced in January, is led by Etzioni and funded by Uber co-founder Garrett Camp through his Camp.org nonprofit foundation.
The organization launched a collection of political deepfake detector AI tools in April, making them available to journalists and fact-checkers. It now has thousands of active users, Etzioni said.
Etzioni said he learned of the shooting in Pennsylvania when he saw that a member of the press from a major news organization was using TrueMedia to check the authenticity of the photograph that appeared to show the bullet in mid-air. He said he was heartened to see the journalist doing their homework in that way.