Experts urge Congress to clamp down on deepfakes

WASHINGTON (CN) — Don’t let the images of Pope Francis in a puffer jacket fool you, a panel of tech experts told members of Congress Wednesday as they warned lawmakers about the profound dangers posed by deepfake technology.

Deepfakes, a term referring to images, audio or video generated by artificial intelligence and designed to depict a real person or event, have become nearly ubiquitous on the Internet in recent years. Some deepfake content is satirical or humorous — TikToks that imagine President Biden playing video games or the notorious image of the Pope wearing a winter coat — but deepfakes can also serve a much more nefarious purpose.

“According to one study, 96% of deepfake videos are nonconsensual pornography,” said Virginia Representative Gerry Connolly during a hearing in the House Oversight Committee’s technology subpanel Wednesday afternoon. “Another report confirms that deepfake pornography almost exclusively targets and harms young women.”

Connolly referenced news last week that students at a New Jersey high school had created sexually explicit images of their classmates using AI tools.

“This is wrong,” he said. “It threatens lives and self esteem among young people and it needs to be stopped.”

Congress has for months explored possible avenues for regulating artificial intelligence, hoping to get out in front of the rapidly growing technology. That effort has developed alongside the White House, which at the end of October issued an executive order laying out new standards for AI safety and security.

The push to better understand artificial intelligence continued Wednesday, as a panel of experts invited to testify implored lawmakers to help mitigate the malicious use of AI-powered deepfakes.

“Despite the president’s executive order and the testimony of our thought leaders and business leaders, we’re not moving fast enough to curtail the continued damage this technology is doing and will do as it evolves,” said David Doermann, chair of the computer science and engineering department at the State University of New York at Buffalo.

Doermann, who previously worked as a project manager at the Defense Advanced Research Projects Agency, cautioned that, in addition to creating nonconsensual pornography and other compromising media of private citizens, deepfakes can be used to impersonate government officials, military or law enforcement.

Not only are people generating such dangerous deepfake content, but it’s already too late to take a targeted approach to fighting it, Doermann said. “Today, this is no longer a problem that can be solved by simply detecting and removing generated content from our social media and content provider sites.”

Groups already at risk of discrimination, such as women, are disproportionately impacted by deepfakes, said Sam Gregory, executive director of the tech-oriented human rights advocacy group WITNESS.

“Women and girls are widely targeted with nonconsensual sexual images,” Gregory said, “and the problem is escalating.”

Deepfake technology is also being used to “muddy the waters” and create confusion around real content online, he added.

Doermann suggested that Congress pass legislation addressing deepfake technology generally. Among its provisions, he said, a good deepfake bill would take steps to improve public awareness and media literacy.

It may seem obvious that people would want to know whether an image they are looking at was generated by artificial intelligence, or that people would take care not to spread misinformation, he said. “But we find that’s not the case.”

Congress must also expand privacy and consent laws to protect people from having their image or likeness used without consent, Doermann added, and lawmakers must fund further research and development to help counter the malicious use of deepfake technology.

Gregory agreed, suggesting that lawmakers dial in on existing harms exacerbated by AI such as nonconsensual sexual content and child sexual material.

“Incorporating broad consultation with groups working on these existing harms while safeguarding constitutional and human rights would help you craft appropriate steps,” he said.

Mounir Ibrahim, a former Department of State foreign service officer who now works as vice president of public affairs at media authentication company Truepic, said that there is “growing consensus” that more transparency is needed online to help users determine whether media is genuine, or AI-generated.

Congress should take steps to ensure that internet users are able to see the “provenance” of media, Gregory said, referring to details of how an image, video or audio clip was made, edited and distributed, among other things. However, he cautioned that implementing provenance standards for human-made content could pose a privacy risk.

“We should be wary of how any provenance approach can be misused for surveillance and stifling freedom of speech,” Gregory warned.

While the witnesses agreed that deepfake technology and artificial intelligence generally poses some real threats, they were pleased with Congress’ attention to the subject.

Ibrahim applauded lawmakers for holding hearings on deepfakes and other artificial intelligence technologies. “Events like this will raise awareness, help educate the public and give an opportunity to ask the right questions,” he said.

“We have created these problems,” Doermann said, “but I have no doubt that if we work together, we are smart enough to figure out how to solve them.”

Follow @BenjaminSWeiss

Subscribe to Closing Arguments

Sign up for new weekly newsletter Closing Arguments to get the latest about ongoing trials, major litigation and hot cases and rulings in courthouses around the U.S. and the world.

Email *

Submit