KQED's Forum Deepfake Videos Just Got More Realistic…and More Dangerous
Oct 8, 2025
Max Read, a seasoned journalist and editor, joins Alice Marwick, a research director focusing on technology's societal impacts, along with Jason Koebler, an investigative journalist, to discuss the alarming rise of AI-generated deepfake videos. They explore how OpenAI’s Sora app makes creating convincing videos effortless and its implications for trust and disinformation. The trio debates the urgent need for regulation, the potential harassments linked to video likeness sharing, and the risks AI poses to political integrity and public trust.
AI Snips
Chapters
Transcript
Episode notes
Instant Realism From A Selfie
- Sora captures a user's face with a short selfie cameo and inserts them into realistic videos in seconds.
- Jason Koebler says this makes deepfakes easy and far better than prior janky models.
Early Adopter Playfulness
- Early adopters enjoy exploring prompts and the novelty of AI-generated video like early social networks.
- Max Read compares the vibe to the first months of ChatGPT or early Facebook adoption.
Seeing Still Equals Belief
- Seeing is still a powerful truth cue and AI video amplifies harm by making false events appear real.
- Alice Marwick warns manipulated video will surpass harms from earlier tools like Photoshop.