The person arguably best equipped to spot AI-generated fakes is now admitting defeat. Hany Farid, a UC Berkeley professor with more than 20 years of experience as the world's leading digital forensics expert, says he is now struggling to identify AI fakes, according to the New York Times.
That's a striking concession. The same tools that once left obvious fingerprints have improved so rapidly that even trained specialists are getting fooled.
The problem is arriving at an especially dangerous moment. According to the Wall Street Journal, AI deepfakes are growing "weirder and harder to spot" ahead of midterm elections, raising fresh alarms about the integrity of the political information environment.
The threat isn't confined to elections. Al Jazeera reports that AI is being weaponized against Muslim women in India, with victims describing fabricated images that "looked so real." The targeting of a specific religious and gender group shows how deepfake technology is being used not only for political manipulation, but for targeted harassment and intimidation of vulnerable communities.
What makes this moment distinct from earlier deepfake scares is the combination of speed and accessibility. The technology has moved from well-resourced actors to tools available to almost anyone — and it is outpacing the detection methods designed to catch it.
If the world's foremost forensics expert can no longer reliably tell real from fake, the implicit trust most people place in photographic or video evidence is no longer warranted — and that has consequences that reach far beyond any single election.