SF firm fights deepfakes as lawmakers target deceptive election videos
Gov. Gavin Newsom just signed three bills into law on Tuesday requiring social media companies to do more to police deep fake videos, especially as we head into this year's election cycle.
It's all in an effort to prevent mis-information from spreading, and one company here in San Francisco is developing new software to detect altered images.
Some deepfake videos are easy to spot, like the spoof Tom Cruise Presidential campaign ad in 2020, or former Gov. Arnold Schwarzenegger appearing in "The Wizard of Oz".
Unfortunately, other deepfake videos aren't as obvious, like a new fake Kamala Harris presidential campaign parody ad.
That's where Hive comes in.
"Usually if you saw a video that looked real, you'd believe what it said, but now that's not the case, right?" said Kevin Guo, the company's CEO.
The San Francisco-based company started to help with content moderation but has since focused on identifying AI-generated pictures and videos. The company uses advanced algorithms and AI to be able to detect the fakes.
"The way our models work don't really interpret the content the way that a human does. It's really looking at kind of on a pixel level basis in a way that's almost invisible to humans, but there is that signature. That watermark there that our models have figured out," Guo told CBS News Bay Area.
Elon Musk re-tweeted that Harris campaign ad parody video, resulting in a Twitter spat with Newsom.
"The federal government, humbly I submit, for many different reasons has failed to regulate. In the absence of that regulatory framework, California asserts itself. We feel a responsibility, particularly as the birthplace of so many of these technologies and so much of this innovation," Newsom said at the Salesforce Dreamforce conference in San Francisco on Tuesday.
While on stage with CEO Mark Benioff, the governor signed three bills into law that require social media companies to label deep fake videos as parody - or remove them completely.
The bills also limit people and companies from sharing deep fakes that spread political misinformation within 120 days of an election.
"The ability for bad actors to kind of influence elections now with widespread misinformation campaigns, this is pretty unprecedented. You can pass a bill that bans political deepfakes, but you still have to be able to detect these deepfakes and kind of sus them out in some way," says Guo.
California Rep. Adam Schiff, who is also running for Senate, introduced bipartisan legislation Tuesday at a federal level to increase regulation around deepfakes. That bill would give the Federal Election Commission the power to regulate the use of AI in campaign ads.
