Deep Fakes a Problem That Will Require Industry, Gov't to Work Together, Speakers Say

“There’s no single silver bullet” to address fake information online, said SIIA President Jeff Joseph. “The threat’s global, large and real, and requires us to use myriad tools in response,” he said. “It requires collaboration between the public and private sectors,” he said: “Technology companies are responding and creating tools. … But it’s a deeper problem.”

“From very early we’ve had an interest in how do you prove what is trustworthy, how do you show what’s been manipulated,” said Sam Gregory, program director at Witness, a human rights group. “There are a whole kaleidoscope of threats that deep fakes … exacerbate,” he said. “That’s why investing early in some of the solutions and thinking about how we address them is really important,” he said. Many governments target their citizens with manipulated images, he said.

Almost all fake images and videos aren’t deep, but really shallow fakes, Gregory said. That’s the phrase people use for a video or photo that purports to show one thing but is really from an earlier date, or a photo altered to remove content, or a video that’s slowed down, he said. “That continues to grow in volume and diversity,” he said. Platforms like TikTok are based on “the idea that you can remix and make it easier and easier to do so,” he said.

“What happens when we move to the metaverse?” asked Hodan Omaar, Center for Data Innovation senior policy analyst. As augmented and virtual reality develop, what happens when deep fakes are created using these new technologies? she asked. The U.S. can lead the way, Omaar said. One of the things the U.S. has been “very good at” on standard setting is “taking this approach of really bringing in academia, having it be industry led, and really engaging every stakeholder that needs to be at the table,” she said: “This is something the U.S. can do well and should do well.”

Adobe studied using artificial intelligence to detect deep fakes, and found it isn’t accurate enough, said Dana Rao, Adobe general counsel and chief trust officer. Detection will get better but so will the ability to create deep fakes, he said. It’s equivalent to a “cat and mouse” game where “one side gets better, and you’re always sort of at a level playing field,” he said. Once people start to think they can’t trust images and videos posted online “you’re not going to believe anything that’s true either, and when we start not believing true things, I don’t know how we function as a society,” Rao said. As one solution, Adobe created software that allows people to click on a box and find out everything that has happened to an image or video after it was captured, he said. Adobe is offering the technology as open-source software, he said.

The Biden administration is focused on promoting democracy, said Tim Maurer, director-technology and democracy at the National Security Council. “There is a global environment and there are some countries who have a different vision for what technology should look like and who controls technology,” he said. “The second pillar is countering the improper use of those technologies,” he said. Digital technologies can promote human rights but also be used to suppress those rights, he said. “The administration has also been very clear that we’re not only going to outline what we think everybody should be doing … but we’re also taking steps to hold actors accountable if they’re not acting how they should and in line with international human rights,” he said.