UC Berkeley Professor Hany Farid, a globally recognized expert in digital forensics, has stated that the race between generative AI and detection tools is practically lost. The technological gap between the speed of creating fakes and the ability to detect them has shrunk to a critical 6–12 months.
What Happened
Hany Farid reported that traditional methods of searching for visual artifacts in generative content (forensics) are becoming ineffective due to the rapid evolution of generative architectures. Under current conditions, deepfake detection methods systematically lag behind the models that generate them, making purely visual verification unreliable.
Context
Given this technological lag, the industry focus is being forced to shift from attempts to "see the fake" toward implementing cryptographic content provenance. This implies a transition to trust architectures where authenticity is confirmed by a digital signature rather than visual analysis.
Why It Matters for the Industry
For the industry, this means a necessary shift from an artifact-hunting strategy to the implementation of standards like C2PA and the creation of "Zero Trust Media" systems, where content without a digital signature is considered untrustworthy by default. Increased legal liability for manufacturers of synthetic content is also expected.
Why It Matters for Users
Ordinary users can no longer rely on their own visual perception when verifying the truthfulness of videos or photos. To obtain reliable information, it is necessary to use institutional sources and tools that support content digital signature verification.
Sources
Author
Look at AI, Editorial Staff