ROB SCHMITZ, HOST:
That is ALL THINGS CONSIDERED from NPR Information. I am Rob Schmitz. Right here in Germany, a narrative of betrayal, intercourse and synthetic intelligence has captivated a nation, and it simply might have implications for all of us. Forty-four-year-old movie star Collien Fernandes has accused her fellow movie star ex-husband, Christian Ulmen, of spreading pictures of her that have been later remodeled by synthetic intelligence into pornographic movies. They have been then circulated over the web. Ulmen denies the allegations. 1000’s of Germans have taken to the streets, demanding stricter legal guidelines for these behind these so-called deepfake movies. However how do you police one thing that is changing into simpler to create and tougher to hint?
We’re joined now by Harvard legislation professor Rebecca Tushnet, codirector of the Berkman Klein Middle for Web and Society. Welcome, Rebecca.
REBECCA TUSHNET: Thanks.
SCHMITZ: So, Rebecca, I really feel like these AI platforms are evolving so rapidly that it is a bit of troublesome to maintain up. How straightforward is it to make use of these platforms to create these deepfake movies utilizing somebody’s picture?
TUSHNET: So the attention-grabbing factor to me is that should you have a look at the timeline of this, the allegations point out that, you recognize, the software program is being utilized in a time when it wasn’t essentially as widespread as it’s now. However this functionality has been accessible on the excessive finish for a very very long time. The previous few “Star Wars” films, Lucasfilm has been ready to do that for some time.
The distinction is there is a proliferation of, like, nudify (ph) apps. So mainly, upon getting the fundamental thought of doing this, it’s attainable to coach a mannequin particularly to create pornographic pictures. And although the massive names that you have heard of – like Anthropic, OpenAI and so forth – have guardrails towards this, it isn’t all that onerous for somebody to go, you recognize, prepare their very own mannequin. So, you recognize, it is very exhausting to place again within the bottle. And there are individuals who can make cash off of it, which is why it is very exhausting to place again within the bottle.
SCHMITZ: So within the case right here in Germany, the allegation includes somebody identified personally to the sufferer. To begin with, how frequent is that? After which how a lot tougher does that make these circumstances to show?
TUSHNET: It is a good query as a result of a whole lot of instances we do not essentially know. Usually, when there’s sufficient proof, it’s attainable to hint again to somebody, you recognize, who might tangentially know the individual – typically like an employer, typically any person that they’d a contact with, you recognize, at a earlier job or a failed date with. But it surely’s typically a shocking individual. I’ve to say the husband is, you recognize, a very shocking individual. But it surely’s as a result of some males really feel an entitlement to do what they need, and the problem is, typically, kind of determining who it was, which may be troublesome, after which, you recognize, attempting to attenuate the hurt attributable to the dissemination.
SCHMITZ: The TAKE IT DOWN Act, which Congress handed in 2024 within the U.S., requires platforms to take away offensive, deepfake content material inside 48 hours after a criticism. I imply, that appears fairly easy, proper?
TUSHNET: Proper. The difficulty is that it does not essentially cease the photographs from shifting elsewhere or, you recognize, from being tweaked a bit of bit and reposted. And I believe everybody understands this. That is to present you a device to cease the bleeding. However frankly, I do not really imagine that felony legislation alone, you recognize, will remedy the issue.
SCHMITZ: So what extra must be executed to guard individuals from this?
TUSHNET: So I believe we face a bunch of various issues. Certainly one of them is only a lack of robust social norms towards this. So…
SCHMITZ: Proper.
TUSHNET: …The analogy that I like to attract is definitely drunk driving within the U.S. So there was a time when drunk driving was not thought-about flawed, even post-criminalization. So what needed to occur was a social shift that claims, you recognize, this isn’t one thing a accountable individual does. And that does not imply that no one drives drunk, proper? It nonetheless occurs. However the dialog round it modified. And I believe that is the type of media literacy schooling, you recognize, anti-bullying, schooling – these are the sorts of steps that may make the long-term change.
However I need to be clear, like, it is nonetheless going to be attainable to do that. In truth, you recognize, traditionally, it was attainable to do that. There have been photo-manipulated, you recognize, pornographic pictures earlier than. What we are able to try to do is get it extra underneath management.
SCHMITZ: Proper. I imply, within the case that – you recognize, the parallel you make to drunk driving, I imply, one of many components in lowering drunk driving additionally have been simply actually harsh penalties for folk who have been caught ingesting and driving. I am questioning, would that kind of legislation work in the USA? I imply, that is what they’re speaking about doing right here in Germany, for instance.
TUSHNET: I am a bit of hesitant about that ‘trigger within the U.S., we are likely to overcriminalize and type of undersocialize. I am guided by that as a result of though this case is – it is clearly horrible – it is uncommon by way of the issues that we’re seeing which might be inflicting essentially the most disruption, which is to teenage ladies. And the individuals who do which might be normally teenage boys. And so I am not normally a supporter of sending a young person to twenty years in jail.
You realize, there’s the explanation why we are saying, you recognize, youngsters aren’t – haven’t got totally developed brains. They will nonetheless do unimaginable hurt, and I do not in any approach need to reduce that. A part of the factor that we do is, you recognize, we now try to train them, for instance, how one can drive. Now we have graduated entry necessities. So I believe we have to do issues like that about accountable web use.
SCHMITZ: Rebecca, that is such a disturbing development, and I am questioning are we seeing extra of this or will we see extra of this as AI platforms proliferate increasingly?
TUSHNET: My sincere guess is that what we’ll proceed to have is a state of affairs the place the well-capitalized, you recognize, distinguished ones – probably aside from Grok – are constructed with guardrails that forestall this type of factor, or at the least attempt very exhausting to forestall this type of factor. However there is a kind of little substrate of scammy (ph) little apps that provide and typically ship the flexibility to do that for people who find themselves keen to go searching.
SCHMITZ: That was Rebecca Tushnet, codirector of the Berkman Klein Middle for Web and Society at Harvard Legislation Faculty. Rebecca, thanks.
TUSHNET: Thanks.
Copyright © 2026 NPR. All rights reserved. Go to our web site phrases of use and permissions pages at www.npr.org for additional info.
Accuracy and availability of NPR transcripts might differ. Transcript textual content could also be revised to appropriate errors or match updates to audio. Audio on npr.org could also be edited after its authentic broadcast or publication. The authoritative report of NPR’s programming is the audio report.
Leave a Reply