Investigating AI Deepfakes

Summary by Krater

Coffeezilla investigates the rise of AI-generated deepfakes, examining their use in scams, political propaganda, and explicit media, and concludes that low-effort, scalable deepfakes pose a severe and under-recognized threat to digital trust.

From the video

Answers: How dangerous are AI deepfakes and how can we tell what is real?

• AI deepfakes
• scams and fraud
• political propaganda
• explicit non-consensual media
• detection technology

What it concludes

• Low-effort, consumer-grade deepfakes are becoming realistic enough to fool many people and pose a far greater scam threat than Hollywood-level deepfakes.
• Human accuracy in detecting AI-generated images, audio, and video is barely above chance (around 50 to 70 percent in lab settings) and full-blown video detection is likely to become obsolete within 12 months.
• AI is enabling single operators to produce sophisticated scams, malware, and political propaganda at scale, making false media a major threat to digital trust.
• Platforms and lawmakers are lagging behind, putting the burden of reporting non-consensual imagery and scams onto the victims rather than the tech companies.