Aug 27, 2026 · 22 min · 9 segments
In this episode, Katherine Forrest and Scott Caravello break down how AI companies are embedding invisible watermarks and provenance signals into the content their models generate. They also walk…
Catherine ForrestHost
Scott CaravelloHostToday, we're going to talk about something that's been actually in the news a fair amount, and it's about transparency in AI-generated content.
How you can tell if content that you're looking at is actually AI generated or not.
And right now, you know, this has become a big issue because there's so many things that I know I look at and my kids say, are you joking? This is, of course, AI generated.
So I'll look at some incredibly cute house cow that's like wandering around and it's like perfectly fluffed and...
Do you see it's got like four ears or whatever? And or something sort of less than that.
But how companies are making it apparent to the viewer when you look at an image or hear a voice or read a paragraph, whether or not that thing that you're viewing is AI generated or not has become a big issue.
So on the engineering side, the big AI companies are now putting what are called hidden signals.
I just want to sort of like alert the audience to the fact that the word signal just means sort of like a marker, an electronic or a digital marker that is apparent in digital content.
So there are companies that put these hidden signals into the content that they generate so that they can actually either follow whether or not certain content is theirs and where it goes, but also so that you can trace back and have sort of a provenance or provenance, as they would say, if it was wine, about where it came from.
And so it also related to all of this sort of the desire to sort of figure out whether or not something is AI generated or not.
There's a now a legal side of things and there's some sweeping new transparency laws that are taking effect both in the United States and in Europe.
And so there are certain places where it's now mandatory that you actually include disclosure that a voice, that an image that has been AI generated.
Today, we're going to talk about something that's been actually in the news a fair amount, and it's about transparency in AI-generated content.
How you can tell if content that you're looking at is actually AI generated or not.
And right now, you know, this has become a big issue because there's so many things that I know I look at and my kids say, are you joking? This is, of course, AI generated.
So I'll look at some incredibly cute house cow that's like wandering around and it's like perfectly fluffed and...
Do you see it's got like four ears or whatever? And or something sort of less than that.
But how companies are making it apparent to the viewer when you look at an image or hear a voice or read a paragraph, whether or not that thing that you're viewing is AI generated or not has become a big issue.
So on the engineering side, the big AI companies are now putting what are called hidden signals.
I just want to sort of like alert the audience to the fact that the word signal just means sort of like a marker, an electronic or a digital marker that is apparent in digital content.
So there are companies that put these hidden signals into the content that they generate so that they can actually either follow whether or not certain content is theirs and where it goes, but also so that you can trace back and have sort of a provenance or provenance, as they would say, if it was wine, about where it came from.
And so it also related to all of this sort of the desire to sort of figure out whether or not something is AI generated or not.
There's a now a legal side of things and there's some sweeping new transparency laws that are taking effect both in the United States and in Europe.
And so there are certain places where it's now mandatory that you actually include disclosure that a voice, that an image that has been AI generated.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.