World's Leading Deepfake Detection Expert Says He No Longer Trusts His Own Eyes
Should tech companies be legally required to embed verifiable authenticity credentials in all AI-generated images, audio and video?
Hany Farid, a UC Berkeley digital forensics professor widely regarded as the field's foremost expert in spotting manipulated images and video, said in a New York Times profile published in June that he can no longer reliably tell real footage from AI-generated fakes. Farid, who has spent decades building detection tools used by law enforcement, journalists and the military, told the Times "I feel like I am going blind," describing how AI video and voice generators have eliminated the glitches, like unnatural blinking or distorted fingers, that once gave deepfakes away. Farid is leaving UC Berkeley's School of Information at the end of June to return to Dartmouth College.
Should tech companies be legally required to embed verifiable authenticity credentials in all AI-generated images, audio and video? The shift comes as deepfake volume has surged. Cybersecurity firm DeepStrike reported deepfake content grew roughly 900% over the past year, and Farid's own research has found that most people, and increasingly even trained analysts, can no longer distinguish real photos, video or voice recordings from AI-made ones. Farid said that even when his team can prove a piece of media is fake, the analysis can take hours, by which point, in his words, "the whole ballgame's basically over" and the false version has already spread.
In response, some technology companies are shifting from after-the-fact detection toward authentication standards like Content Credentials, a digital record embedded in a file at creation showing how it was made and whether it was altered. Reported deepfake scams already include fraudulent emergency phone calls that clone a family member's voice, fake celebrity endorsements, and financial fraud schemes using AI-generated executives.
What supporters say:
Supporters say voluntary standards alone won't scale fast enough given deepfake content is growing by roughly 900% a year, so embedding provenance data by law is the only way to keep pace.
Backers argue this shifts the burden from individuals like Farid, who says thorough detection now takes hours per file, onto the platforms and generation tools that created the problem.
Advocates note authentication standards don't ban AI content, they just label its origin, preserving legitimate uses like satire, film and marketing while flagging deceptive ones.
What critics say:
Critics warn mandatory watermarking or credentialing schemes are easy to strip out, since platforms like X already remove metadata to reduce file sizes, undercutting any legal requirement in practice.
Opponents argue open-source AI models, which anyone can download and run on consumer hardware, sit outside any single company's control, making a company-focused mandate incomplete.
Some in the industry say the better fix is procedural, like family "safe words" and callback verification, rather than a technical mandate that gives people false confidence in a system bad actors can still evade.
What's your take?
Should tech companies be legally required to embed verifiable authenticity credentials in all AI-generated images, audio and video? Yes 
No 
Other 
Sources:
#AI #Deepfakes #TechPolicy #DigitalForensics #Misinformation
Now let's hear from you.
You get one Take and 3 ratings so use them well.
Strong arguments beat loud ones.
Don't forget to rate your own Take.
Please don't feed the trolls.
