World’s Best Deepfake Expert Can’t Detect AI Videos

Hany Farid has declared that artificial intelligence image generation has reached a point where it can fool even the world’s greatest experts, including himself.

Farid is a professor at the UC Berkeley School of Information and the senior faculty adviser for the Center for Long-Term Cybersecurity. He is widely considered to be the world’s leader in digital forensics and image analysis. But according to a report from The New York Times, Farid is struggling to pick up on some AI-generated content.  

Farid has noticed that AI has improved dramatically at mimicking real shadows, projectile trajectories, plant life, human faces, and sound, removing many of the clues that digital forensics experts relied on to detect deepfakes.

“I don’t trust anything. Every image I see, I’m drawing lines for shadows and doing geometry in my head, trying to figure out what I’m looking at. It’s over,” Farid said. “Within a year or two, our whole visual system will be utterly useless.”

Farid has collaborated with Microsoft to create PhotoDNA, software used by many major tech firms to detect child sexual abuse material. Additionally, Farid worked with the U.S. military to create scanning software to detect deepfake videos before the 2020 election, according to an article originally published in The Los Angeles Times.

Farid told The New York Times that he is frustrated with how Big Tech is rapidly developing AI with no concern for the repercussions and hopes to retire from the scene and move to a farm in Vermont.

“I can’t stand this place anymore,” Farid said. “These major tech giants will burn everything to the ground as long as they’re making a profit. They’re not interested in anything that’s going to slow them down.”

According to cybersecurity firm DeepStrike, deepfake files have increased dramatically from 500,000 in 2023 to 8 million in 2025, coinciding with a 3,000% increase in fraud attempts worldwide.

Now that human detection systems are failing, experts are fighting fire with fire by using AI to combat AI deepfakes, such as the AI-powered Deepfake-O-Meter and Farid’s own company, GetReal Security.

Additionally, experts are advocating for new standards for digital photography to make it easier to determine a photo’s authenticity. 

For example, the Content Credentials movement is advocating for a pin to accompany digital photos that will allow viewers to “determine the method of creation” for a photo “and see a record of editing history.”

Copyleaks, an AI content-detection platform, has detailed five major emerging deepfake scams potentially jeopardizing Americans’ well-being.

Celebrity scams utilize deepfakes of famous people to promote products, while catfishing scams involve deepfaking a beautiful person to manipulate lovers into giving money or personal information. 

Fake security footage is being used to prosecute innocent people, and scammers are using AI voice changers to make people believe a loved one has an emergency and is asking for help. 

In the political realm, scammers are making fake videos of political candidates making controversial claims they never actually made.