Deepfake technology is becoming more convincing — and more dangerous — as artificial intelligence improves.
They are no longer limited to prerecorded videos. Deepfakes can now be used in real time.
“There are apps on the internet people can actually access, which can create real-time impersonation of someone, especially in online interview or meeting settings,” said Professor Siwei Lyu, director of the University at Buffalo Institute for Artificial and Data Science.
Lyu said triple-checking whether someone online is a real person will likely become standard practice as deepfakes continue to improve.
“You literally can put somebody in a totally different background based on maybe one image of that person's face, create a full animation of that person,” he said. “And the quality is very high, that it's difficult to tell them apart.”
Face-swap deepfakes surged more than 700% in 2023, and by 2024, a deepfake fraud attempt occurred every five minutes, according to identity security firms.
Deepfake experts like Lyu say detection tricks can still be useful.
“Can you take your hand and put it in front of your face to like cover it partially?” Lyu said.
But he noted those tricks will likely be addressed as the technology improves.
“The technology is developing very fast, but it has not reached the level that we truly have no way to tell if this is a real person or not yet,” Lyu said.
Deepfakes are not going away, but experts say making them harder and more expensive to carry out could help slow their spread and limit the most harmful uses.
