An ”Is this really legitimate or is it AI?” moment

I just had a reminder of the possibilities and perils of the evolving media environment.

I often quote the leading “realist” foreign policy scholar John Mearsheimer here, usually at least once a week.

Just in the last few days, I noticed two new YouTube channels featuring Mearsheimer: “Mearsheimer Lens” and “Mearsheimer Realism.” Both appear to be technically well-done AI productions. The former describes itself as follows:

The Mearsheimer Lens is a fan-made channel dedicated to exploring global politics through the lens of Professor John J. Mearsheimer’s realist perspective. We are not affiliated with or representing Professor Mearsheimer in any official capacity — this channel simply aims to discuss and interpret his ideas on international relations, power politics, and world order.

… Inspired by John J. Mearsheimer - not speaking for him. (1)

But as of this writing, it doesn’t say the images and sound it shows are AI-generated or excerpts culled from other interviews. Today I listened to a very recent one about Venezuela, which included AI-Mearsheimer talking about how he thinks Venezuela’s cooperation with Russia and Iran fits into his “offensive realist” foreign policy view. And it sounded pretty good, and pretty Mearsheimer-ish.

But I checked again to see if I could find any other additional information on the source. There’s the description quoted above that disavows any claim of “speaking for” him and so forth. “Mearsheimer Realism” has been online only since earlier this month. “The Mearsheimer Lens” has been around since 2015 but I haven’t noticed it until recently. (AI’s capabilities in 2015 weren’t anything like they are today in creating credible digital clones of people.)

One thing that seemed unusual in the Venezuela-related video is that AI-Mearsheimer is speaking directly to the camera at length, with a backdrop that appears in many of his legit interviews. (Or at least a lot like it.) He typically sits for interviews with a live interviewer – a with various different kinds of interviews, from grumpy conservative isolationists to Global South lefties. He seems to find that format to be a good one for him, and he seems to appreciate probing questions.

I’ve also seen videos of him giving lectures, where he talks at some length on his own. But AI-Mearsheimer speaking on his own sounded mostly like the real one does when he is responding to interview questions and not so much like he was giving a lecture. He also seemed more repetitive than real-Mearsheimer normally is. And several times, the pronounced words ending in “-tion” with a bit of a Spanish-language accent, which I’ve never noticed with real-Mearsheimer.

I also did come across a website (that I had never heard of before) called Jammerjoh that claimed that “Mearsheimer Lens” was using an AI-Mearsheimer.

Evolving technologies don’t go “back into the box.” AI has been around for a while -

But consumers of its output also have to adapt.

So do regulators. I was able to piece together the above because I’ve been a geeky blogger since 2003 and I’ve always tried to be careful about sourcing. I really like the footnoting capability of Substack, not least because back in the early days of blogging, it was common to use hyperlinks for sources. For a while, I even saw people criticize bloggers from quoting text from another blog but without including the hyperlinks from the original in the quote.

One problem then was that hyperlinks were not nearly as stable as they are now. I remember once spending at least an hour, maybe more, trying to track down a New York Times article I had quoted in a previous post just a few days before, because the hyperlink I had used had already died. So I started early on showing the titles of articles and the sources including a hyperlink for the title.

Between that and my academic training, I try to vet my sources carefully.

Back in the days when there were only three major TV networks in the US, only a few newspapers like the New York Times and the Washington Post could be considered national and/or international papers. And they actually were, you know, on paper. There were national monthly and weekly magazines. Time, Newsweek, and the stodgily conservative U.S. News and World Report were the three dominant ones in the US. And you had to get them from either newsstands (there are still some of those with a kind of nostalgic charm now), by mail subscription, or at a library.

And the only people then who had phones they could hold in the palms of their hands were characters on Star Trek.

So then, it was a bit easier to know where to go to do reality-checks on particular factual claims from current news reports. Not so easy to get to them, maybe, but a more limited number of places to go to look. This had its downsides, too. The famous book by Edward Herman und Noam Chomsky, Manufacturing Consent, about how governments and economic elites could massively influence public opinion even through the independent major media outlets at the time, was first published in 1988. Before everybody had an e-mail account. A The year before Quantum Computer Services changed its name to America Online.

But this doesn’t mean that governments can’t establish reasonable regulations for digital media. For instance, a portal like YouTube could require users who post videos to clearly and prominently state who AI is being used in the video, particularly when they are using the names and images of real people in the AI simulation.

Notes:

(1) Mearsheimer Lens YouTube channel: Description. <https://www.youtube.com/@TheMearsheimer> (Accessed: 2025-31-10).