Hany Farid Warns Deepfakes Are Harder to Spot, Urges Secret Family Passcodes

Generative AI can now mimic real people’s images and voices, making video footage, tone of voice and real-time reactions insufficient to verify someone’s identity. An experiment in which BBC journalist Thomas Germain called friends and relatives showed that even a real person may struggle to prove they are not AI. This “liar’s dividend” also makes authentic evidence easier to dismiss, undermining social trust and digital identity verification.

Digital forensics expert Hany Farid recently warned that clues used to identify deepfakes are becoming increasingly unreliable. He recommended that families and friends agree in advance on a secret passcode known only to them. When faced with a suspicious call, video chat or urgent request to transfer money, people can ask the other party for the passcode as a low-tech defense against AI impersonation scams. The report did not specify a date or any amount involved.