On Point with Meghna Chakrabarti Can we even trust videos anymore?
Dec 15, 2025
Hani Farid, a UC Berkeley professor and digital forensics expert, explores the alarming rise of AI-generated deepfakes, facilitated by the Sora 2 app. He discusses how these tools can create indistinguishable manipulated media, eroding our trust in visual content. Hani emphasizes the importance of understanding how deepfakes work and criticizes tech companies for rushing to release such technology without adequate safeguards. He highlights the broader societal impacts, including threats to democracy and the challenges of regulating this rapidly evolving landscape.
AI Snips
Chapters
Books
Transcript
Episode notes
Deepfakes Are Crossing The Uncanny Valley
- Deepfakes are already indistinguishable: images and voices fool people and video is quickly following.
- AI trained on massive scraped internet data now creates realistic space-time video volumes.
Narratives Amplify Fake Media
- AI-generated media spreads faster when it matches preexisting narratives.
- Minimal realism is enough if content fills what people already want to believe.
Live Zoom Face Swap Demonstration
- During a Zoom demo Hani Farid switched his on-screen face into Sam Altman’s likeness.
- The instant change startled the producer and showed real-time face-swap risks.
