What is a deepfake?

A deepfake is fake media made with AI that looks or sounds real: a video of a politician saying something they never said, a photo of an event that never happened, or a phone call in a loved one's cloned voice. The name comes from "deep learning," the kind of AI used to make them.

How they're made

• Learning a face. The software studies hundreds or thousands of images of a person, then learns to rebuild and swap that face, according to the US Government Accountability Office (GAO).
• Two AIs competing. A method called a GAN pits one AI that makes fakes against another that tries to catch them, so the fakes keep improving.
• Cloning a voice. The FTC says a scammer may need only a short audio clip, easily taken from videos posted online.

Today's generative AI tools have made all of this far easier and cheaper.

The harms

• Scams. Criminals use cloned voices to pose as a relative in a crisis and demand money fast. The FBI and FTC have both warned about this.
• Non-consensual imagery. GAO found much deepfake content online is pornographic and overwhelmingly targets women.
• Elections. In January 2024, robocalls with a cloned voice of President Biden told New Hampshire voters to skip the primary. The FCC proposed a $6 million fine against the consultant behind them.

How to spot and fight them

• Check the face. MIT's Detect Fakes project suggests watching skin texture, shadows around the eyes, glare on glasses, blinking and lip movement. The FBI adds warped hands and odd teeth.
• Verify, don't trust the voice. Call the person back on a number you know.
• Content credentials. The C2PA standard, backed by Adobe, Google, Microsoft, OpenAI and others, attaches a record of where media came from and how it was edited.
• Laws. US enforcement of the TAKE IT DOWN Act began in May 2026, requiring platforms to remove reported intimate deepfakes within 48 hours. The EU's AI Act requires deepfakes to be clearly labeled, with those rules applying from August 2026.