Really? no, Really? with Jason Alexander & Peter Tilden Hany Farid on the Dangerous and Deceptive Deep Fakes | Really? no Really?
Aug 19, 2026
Hany Farid, a digital forensics professor who builds tools to detect synthetic media. He explains how social platforms amplify AI-made deepfakes and why detection is getting harder. He discusses threats to courts and evidence, fraud via fake interviews and voice clones, and defenses like watermarks, authentication, and mannerism models.
AI Snips
Chapters
Transcript
Episode notes
Minimal Data Can Steal Your Likeness
- One image plus ~10 seconds of voice is enough to fully co-opt someone's likeness.
- Hany Farid warns modern generative models only need minimal data to produce convincing deepfakes that mimic your face and voice.
Social Media Fuels Deepfake Harm
- Deepfakes are amplified by social platforms — creation alone isn't the crisis.
- Hany calls the cycle creation→distribution→amplification→monetization a perfect storm that lowered barriers to entry.
Speed Mismatch Breaks Forensic Value
- Verification is often too slow: forensics experts can debunk fakes but only after the internet has moved on.
- Hany's team spends hours fact-checking while false content spreads and becomes postmortem commentary.