Detroit Free Press: What will U.S. do to combat deepfakes? Michigan members of Congress want to know

WASHINGTON — Imagine a scenario where, on the eve of next year’s presidential election, the Democratic nominee appears in a video where he or she endorses President Trump.

Now, imagine it the other way around. Or that both endorse the other.

Preposterous, right?

But it’s exactly what a British think tank, Future Advocacy, was able to show is possible — as a warning of sorts — in the weeks leading up to the Dec. 12 election in the United Kingdom. It created two videos which, despite never happening, appeared entirely authentic, one with British Prime Minister Boris Johnson endorsing his opponent, Jeremy Corbyn; and another, with Corbyn endorsing Johnson.

It’s an example of what are known as “deepfakes,” advanced technologies which allow for video and audio materials to be manipulated and re-created in such a way as to make the unreal real — or at least appear to be so. And in Congress, two Michigan legislators are among those trying to wrap their heads around how best to combat it.

U.S. Sen. Gary Peters, D-Mich., and U.S. Rep. Haley Stevens, D-Rochester Hills, have both backed bills aimed at prodding the U.S. government to help root out deepfakes. The Johnson and Corbyn videos were never aired as being genuine but hint at the sorts of disruption such technology could cause.

With clear evidence that Russian-backed operatives interfered with the 2016 presidential election in the U.S., those threats are even higher going into 2020 as well.

“We already know the potential for manipulation exists in social media platforms and digital media,” said Peters, the top ranking Democrat on the Senate Homeland Security and Governmental Affairs Committee, referring to evidence of bots and fake accounts on Twitter, Facebook and other social media that can often be used to attempt to confuse or coerce viewers into believing one thing or another. “Deepfakes take that to a whole new level.”

By using advanced technology, precise video and audio editing equipment, specialized software, impersonators and AI — artificial intelligence, through which computing technology can review and fix mistakes, learning not to make the same mistakes again — creators of deepfakes can produce material that makes it seem like something happened that never did.

Peters makes the point in a video of his own, in which he seems to voice support for — horrors — Ohio State ahead of its annual game with the University of Michigan, which is something he assures the viewer he would never do. And in 2018, comedian Jordan Peele voiced a deepfake of Barack Obama as a public service announcement about the dangers of such material.

The prospect of sowing chaos and confusion isn’t limited to politics either. Consider:

  • Say a video is created suggesting a military threat, or a biological one, like an infectious disease, in a given area. Peters notes that in 2014, when there were widespread media reports of the Ebola virus, “people stopped going out,” despite the fact the risk of contagion was small. Such a deepfake, if believed, could hurt whole economies or, in a worst-case scenario, cause an armed response.
  • There have been reports of fake pornographic videos being distributed in which the face of one woman is seamlessly superimposed on another woman’s body, appearing to be far more genuine than such efforts in the past. It’s also apparently not just celebrities coming in for this sort of harassment. According to the Washington Post, a ”number of deepfakes target women far from the public eye” as the price of such technology comes down and applications allowing for such fakery become more readily available.
  • Some of the most significant concerns about deepfakes come down to personal security and safety, with worries that technological advances and an increasingly interconnected network of devices could lead, say, to someone telling your computerized home assistant to open the doors to your house to let a stranger in, or make a purchase on someone else’s behalf. In a related kind of case, a hacker recently managed to get onto a Mississippi family’s home network and, through a camera and microphone installed there, talk to an 8-year-old girl, making racial slurs and asking her to misbehave.

Part of the problem at present, however, is that the technological, scientific and legal communities are not as far along as they need to be in determining what is a deepfake and what is not and how to address it. That’s where Peters’ and Stevens’ pieces of legislation come in: Peters’ bill, which he co-sponsored with Sen. Rob Portman, R-Ohio, and has passed the Senate, would push the Department of Homeland Security to do annual reports on the state of “digital content forgery” and how to combat it.

Stevens, meanwhile, who is chairwoman of the Research & Technology Subcommittee in the House, co-sponsored a bill which has passed the House that directs the National Science Foundation to begin committing funding to learning all it can about detecting and stopping “generative adversarial networks” — otherwise known as GANs — the underlying computing networks which help to create deepfakes.

“What we know about deepfakes is we don’t know a lot,” said Stevens. “When I talk to the social media companies, they’ve started to embrace a set of procedures that put in place some standards, some flags for things that appear to be manipulative or are not true. We could do a lot more.”

To read the full article, click here.