Viral footage of software being used to create AI-generated deepfakes in real time over live webcam feeds has given rise to new fears about the potential use of this artificial intelligence (AI) technology for everything from financial fraud to election interference. The Deep-Live-Cam software can extract a face from a single photo and apply it to another person on a webcam feed live.
Videos showing imitations of Republican Vice Presidential candidate J.D. Vance, Elon Musk, Mark Zuckerberg, and actors George Clooney and Hugh Grant in real time have driven interest in the open source software. They also demonstrate the ease with which deep fakes can be created thanks to developments in the technology.
Risks of AI Deepfake Technology
According to Ars Technica, deepfakes have already led to several successful high profile incidents of fraud. In one, someone stole more than $25 million dollars from a Hong Kong company after impersonating its CFO on a video call.
The Deep-Live-Cam software compiles multiple software packages in a single interface that detects faces in the source and target images and uses the “inswapper” and GFPGAN AI models to swap faces and enhance the footage. Though it’s not yet ready for widespread use, and is not as easy as plug-and-play installation, the software package puts sophisticated technology in the hands of a larger group of users.
The nature of opens source AI software means it is likely to continue to improve as more people use and improve upon it.
Learn more about how to prevent AI-based identity fraud in our eWeek video interview with Ping Identity’s Patrick Harding.