The 2024 US presidential campaign has seen a number of notable deepfakes: AI-powered impersonations of candidates that attempted to mislead voters or humiliate the targeted candidates. Thanks to Elon Musk’s retweet, one of those deepfakes has been viewed more than 143 times million time.

The prospect of unscrupulous campaigns or foreign adversaries using artificial intelligence to influence voters has alarmed researchers and officials across the country, who say AI-generated and manipulated media is quickly spreading online. For example, researchers at Clemson University found an influence campaign on the social platform X that uses AI to generate comments from more than 680 bot-powered accounts supporting former President Trump and other Republican candidates; the network has posted more than 130,000 comments since March.

To strengthen its defenses against doctored images, Yahoo News — one of the most popular online news sites, which attracts more than 190 million visits per month, according to Similarweb.com — announced Wednesday that it is integrating deepfake image detection technology from cybersecurity company McAfee. The technology reviews images submitted by Yahoo news contributors and flags those likely to be AI-generated or faked, allowing the site’s editorial standards team to decide whether to publish them.

Matt Sanchez, president and general manager of Yahoo Home Ecosystem, said the company is just trying to stay one step ahead of the cheaters.

“While deepfake images are not an issue on Yahoo News today, this tool from McAfee helps us be proactive as we always work to ensure a quality experience,” Sanchez said in an email. “This collaboration strengthens our existing efforts and gives us greater accuracy, speed and scale.”

Sanchez said media outlets across the news industry are thinking about the threat of deepfakes — “not because it’s a widespread problem today, but because the possibility of abuse is on the horizon.”

However, thanks to easy-to-use AI tools, deepfakes have proliferated so much that 40% of high school students surveyed in August said they had heard of some form of deepfake footage being shared at their school. The online database of political deepfakes being compiled by three academics from Purdue University contains nearly 700 entries, including more than 275 this year alone.

Steve Grobman, McAfee’s chief technology officer and executive vice president, said the partnership with Yahoo News grew out of McAfee’s work on products to help consumers detect deepfakes on their computers. The company realized that the technology it developed to flag potential AI-generated images could be useful for a news site, especially one like Yahoo, which combines the work of its own journalists with content from other sources.

McAfee’s technology adds to the “rich set of capabilities” Yahoo already had to verify the integrity of material coming from its sources, Grobman said. The deepfake detection tool, itself powered by AI, examines images for the kinds of artifacts that AI-powered tools leave behind among the millions of data points within a digital image.

“One of the nice things about AI is that you don’t have to tell the model what to pay attention to. The model finds out what you should pay attention to,” says Grobman.

“The quality of counterfeits is growing rapidly, and part of our partnership is just trying to stay on top of it,” he said. That means monitoring the state of the art in image generation and using new examples to improve McAfee’s detection technology.

Nicos Vekiarides, CEO of fraud prevention company Attestiv, said it is an arms race between companies like his and companies that make AI-powered image generators. ‘They are getting better and better. The deviations are getting smaller,” Vekiarides said. And while there is growing support among the industry’s major players for inserting watermarks into AI-generated material, the bad actors won’t play by those rules, he said.

According to him, deepfake political ads and other fake material broadcast to a broad audience will not have much effect because “they will be debunked quite quickly.” What’s likely more damaging, he said, are the deepfakes pushed by influencers to their followers or passed from individual to individual.

Daniel Kang, an assistant professor of computer science at the University of Illinois Urbana-Champaign and an expert on deepfake detection, warned that no AI detection tool today is good enough to deter a highly motivated and well-resourced attacker. to capture, like a statesman. sponsored deepfake creator. Because there are so many ways to manipulate an image, an attacker can “turn more knobs than there are stars in the universe to try to bypass the detection mechanisms,” he said.

But many deepfakes don’t come from highly sophisticated attackers. Therefore, Kang said he is positive about current technologies for detecting AI-generated media, even if they cannot identify everything. Adding AI-powered tools to sites allows the tools to learn now and get better over time, just as spam filters do, Kang said.

They are not a panacea, he said; they must be combined with other safeguards against manipulated content. Still, Kang said, “I think there is good technology we can use, and it will get better over time.”

Vekiarides said the public has prepared for the wave of deepfakes by accepting the widespread use of image manipulation tools, such as the photo editors who virtually airbrush the imperfections of photos on magazine covers. The jump from a fake background in a Zoom call to a deepfaked image of the person you meet online isn’t that big, he said.

“We let the cat out of the bag,” Vekiarides said, “and it’s hard to put it back in.”